Results

We first considered the impact of a reduction in regional transportation costs on overall economic welfare. The results of our simulations, using the household EV measure, are presented in Table 4 [ PDF 25.1KB | 1 page ]. The overall welfare estimates are the sums in the row labeled “total”. This type of estimate of the benefit/cost of the proposed change is sometimes called a “one-off” gain/loss. However, this is somewhat misleading as the changes are permanent. Rather we can think of this (roughly) as a permanent increment to household incomes, at constant prices. In absolute terms, the biggest beneficiary in either scenario among the SASEC economies is India, followed by Nepal, then Bangladesh. The total welfare gain is positive, if modest, for all of the SASEC members.

When none of the benefits of transportation cost reductions is assumed to be passed on to other South Asian economies (Scenario 1), both Sri Lanka and Pakistan are estimated to suffer decreases in welfare. However, the losses are small. This suggests the non-SASEC economies in the region are likely to be affected only marginally, if at all, by investments in SASEC transport infrastructure should they be unable to utilize the networks. In the somewhat more reasonable case where some access benefit accrues to the other economies in the region (Scenario 2), the total welfare impact to both Pakistan and Sri Lanka becomes positive, although the benefits to Sri Lanka remain small in absolute terms.

In terms of a true “one-off” measure of benefit/cost, we need to discount the permanent income stream measured by the EV. If the discount rate is assumed to be a standard 2%, then the total estimated benefit is 50 times the annual increment; in Table 4 this row is labeled “cumulative”. We can think of this as the total benefit of the reduction in transportation costs, and this is the figure with which the initial cost of the project needs to be compared to evaluate the net benefit.

It is important to note that while reduction in transportation margins is similar in many respects to lowering a tariff, a CGE simulation of the latter will include all of the costs and benefits, and hence the welfare impact can be directly interpreted as net benefit.

In the case of a reduction in transportation costs that comes via investment in infrastructure, the cost of the investment must also be considered. A CGE model is not designed to estimate the cost of investment, but rather to show how that investment might impact the economic system. Hence, we leave the question of what the proposals outlined in the RETA would cost to experts in project financing. However, we can say from a cost-benefit perspective, according to our topline results, that the project would be worthwhile for India in the aggregate if the investment costs for India do not exceed US$4.3 billion. However, for a variety of reasons, this estimate is likely to be very much a lower bound since the comparative static simulation technique used here does not capture any potential dynamic accumulation effects (i.e., some proportion of the increment to income might be invested, leading to a multiplier effect), and the competitive model used does not account for potential scale effects. Moreover, there could be internal transportation margin effects that a CGE model of this type is unable to capture.

In terms of relative benefits, we can evaluate the estimated welfare impact relative to a baseline metric, most commonly the initial GDP. The final row of Table 4 expresses the cumulative gain as a proportion of GDP. Viewed from this perspective, by far the biggest beneficiary of the reduction in transport margins in SASEC is not India but Nepal, with a cumulative gain of over 12% of GDP in both scenarios. Nepal is followed by Bangladesh, with the gains to India being quite small when expressed as a percentage of GDP. Overall, the results suggest that although the absolute benefits are relatively evenly spread across the members of SASEC, the poorer economies, especially Nepal, benefit disproportionately, relative to their economic size.

In summary, all SASEC member economies are estimated to benefit from the margin reductions, as are non-SASEC economies in the region under Scenario 2. The largest welfare gains in an absolute sense accrue to India, followed by Bangladesh (Pakistan in Scenario 2), and then Nepal. But, this is largely a reflection of the size of the economies in question. Measured relative to GDP, the biggest winner by a substantial margin is Nepal.

Before turning to the estimated impact on household welfare, it is useful to review the household categories in the model, as presented in Table 2.

In the Sri Lankan data, we have five household groups, broken down by location and income level into rural/urban and high income/low income groupings. For Nepal, we are limited to four groups, three rural groups organized by the size of their land holdings, and a single urban group.

The data for India, Pakistan and Bangladesh are grouped by archetype. In India, group H2 (rural agricultural labor) is the poorest group by a substantial margin, followed by H4 (other rural) and H3 (rural non-agricultural labor). The richest groups are H6 (urban self-employed) and H7 (urban salaried). The households differ substantially in their ownership of productive factors, with the richest rural group (H1, rural self-employed) being substantial owners of land and capital. On the other hand, the poorer households, especially H2, receive income almost exclusively from selling their own labor (a large fraction of which is unskilled). Comparing the poorest two groups (H2 and H4) with the richest two (H6 and H7), we observe significant differences in spending patterns as well, although the differences are not as great as in ownership of productive resources. In particular, the two poorest groups spend nearly 2.5 times as much of their income on basic food items (in particular processed rice), as the two richest groups. For textiles the pattern is less dramatic, but the poor groups spend about 30% more than the rich groups.

In Bangladesh, the poorest groups are H1 and H2, rural groups with only limited or no land holdings. They are followed by H7, H3, and to a lesser extent H8, that is, the urban illiterate and poorly educated, and rural households with small land holdings. The richest groups, by a substantial margin, are urban households with high or medium education (H9 and H10). The factor allocation pattern is similar to India, with the lower income groups having a much higher dependence on unskilled labor. Consumption differences are also similar, with the poorest households spending more than double the proportion of their budget on processed rice compared with the richest households.

In the Pakistan data, as with Sri Lanka, we have a combined archetype and income level classification. The data are very detailed, with a concentration on rural households. Households are grouped into multiple farm sizes based on land holdings, and three regions, in addition to the rural rich, and urban poor/rich. In total, our model tracks changes in the behavior of 47 household groups in the region.

The decomposition of the total welfare impacts on the various household groups is summarized in Table 4. Figures drawn in a box are not robust to changes in the underlying parameters of the model, i.e., we cannot be sure of the sign of the change. Other values are robust given the assumptions on the parameter distributions.¹²

Consider first the member economies of SASEC, which are our primary interest. In Nepal, the results indicate that all household groups would benefit, and the results for all households are robust. Of note is that the biggest gainers are small farm households and landless rural groups, while smaller gains accrue to large farm households and the urban group (which is richer on average). Hence, reduction of international transportation margins in SASEC would be pro-poor in Nepal in both an absolute and a relative sense, and also likely be relatively uncontroversial given the uniform benefits. The results follow a similar pattern, but are slightly larger when other economies in South Asia are included in the analysis (Scenario 2).

In Bangladesh, the reduction in transport margins has a positive impact on the welfare of all household groups except H3 and H4. H3 (small farmers) is one of the poorer groups in the country, but the negative impact is relatively small. On the other hand, the loss to relatively large (and relatively rich) farmers is more substantial. The poorest groups (H1, H2, and H7, corresponding to the rural landless, marginal farmers, and the urban illiterate) all experience income rises. This suggests that a reduction in SASEC transportation margins would be pro-poor in Bangladesh in an absolute sense. However, by far the largest gains accrue to H10, the urban highly educated. This is the richest group. Hence, it seems likely that the changes would not lower relative poverty (i.e., income inequality) in Bangladesh. As with Nepal, the pattern is unchanged, although the totals are larger in Scenario 2.

As noted above, the largest absolute gains are estimated for India in both our scenarios. However, India is also the country where the distributional consequences may be most severe. Welfare is estimated to fall in household groups H2, H4, H6, H7, H8, and H9, although the results for H6 and H7 are robust only in Scenario 2. This may be problematic as the poorest groups are H2 and H4. Of the three poorest archetypes, only H3 (rural non-agricultural labor) sees a modest income increase. By far the group that gains the most is H1, large farmers, who are middle income. This suggests that positive changes in the value of agricultural land is the primary driving factor of household income changes in India. Overall, the policy may be marginally pro-poor in a relative sense, as the welfare of the richest groups falls, but probably not in an absolute sense. In reality the reduction in transport margins is strongly pro-agricultural landowners. The same pattern exists and is more pronounced in Scenario 2.

Now consider the non-SASEC economies of South Asia. As noted above, in Scenario 1, the overall welfare impacts of the simulated changes in transport margins are small. The impacts at the household level are small also, and in many cases not robust. The only moderately large impact is a decline in welfare in Pakistan for group H18, the urban rich. In Scenario 2, however, the impacts on the households in Pakistan and Sri Lanka are much more pronounced, as we would expect. In Sri Lanka there are uniform gains across all household categories, although the result for H2 (rural low income) is sensitive to the parameters of the model. For Pakistan, the results are also uniformly positive, suggesting a fall in absolute poverty levels. By far the group that gains the most in Pakistan is H18, the urban rich. Hence, relative poverty may increase.

Overall, the impacts of the changes at the household level exhibit more variation than the aggregate results. While the policy appears to be pro-poor in an absolute sense in many cases (e.g., Nepal), income inequality also seems likely to rise in several cases (e.g., Bangladesh). Moreover, in some cases the poorest groups in society are disadvantaged (e.g., in India). Note, however, that the total welfare gains are positive for all regions in the model, hence it is possible to ensure that all household groups benefit. Our calculations are based on the assumption of invariant transfers, taxes, and factor ownership, but in principle these can be changed if the political will exists.

In addition to overall welfare effects, and their distribution across various groups in the societies in question, CGE simulation also generates information on sectors. Of particular interest are changes in the production structure, both because they indicate which sectors are most likely to be impacted by the proposed policy, and because they provide an indication of the potential degree of structural adjustment required. Estimates of the sector production changes are presented in Table 5 [ PDF 24.7KB | 1 page ]. Again, results that are not considered robust under our sensitivity analysis are highlighted with a box.

Considering the SASEC economies first, in India, the production changes associated with the infrastructure scenario are very small (of course, regional variations may be hidden by the aggregation). The largest changes are in textiles and apparel, which experience very slight declines in both scenarios. The results suggest that for India overall, improvements to transport infrastructure of the magnitude considered here are not likely to cause any significant adjustment difficulties.

In Bangladesh, the production shifts are generally orders of magnitudes higher than in India, reflecting the larger impact of the simulated shocks on a smaller and less diversified economic system. They remain small, however. In Scenario 1 the largest drops in output are in the region of 0.5%. Contractions are observed in heavy manufacturing with the exception of chemicals and rubber, while expansions are observed in light manufacturing (textiles, apparel and leather products). The same broad patterns are observed in Scenario 2, but are more pronounced. However, no single sector expands or contracts more than 1% even in this expanded scenario. This suggests that adjustment issues are not likely to be significant in Bangladesh, either.

In Nepal, the production shifts are larger still, again reflecting the smaller economic system. The largest expansions/contractions exceed 1%. Nepal also seems to be more affected by the assumption of lowered transport costs for trade with Sri Lanka and Pakistan (i.e., comparing Scenarios 1 and 2) in terms of both the magnitude of impact and the pattern. In Scenario 1 we observe significant expansions in textiles, wood products and chemicals, and contractions in apparel and heavy manufacturing. By contrast, in Scenario 2 we observe contractions in textiles, but a smaller contraction in apparel. The contraction in heavy manufacturing is also modulated, but there is a relatively large contraction of metals and minerals. The difference in the outcomes presumably reflects much stronger trade ties between Nepal and Pakistan than between Pakistan and the other SASEC economies. The results suggest that Nepal is likely to face greater adjustment problems than the other SASEC economies, although the shifts remain relatively small overall.

For Pakistan and Sri Lanka the output changes are very small in Scenario 1, as we might expect given that they are impacted only indirectly. The only exception is a moderate decline in metal products in Pakistan. In Scenario 2 the changes are more significant in both economies. In Pakistan we project contractions in light manufacturing and metal products, with only the latter exceeding 1%. Minor expansions are predicted in food products, textiles, and metals and minerals. In Sri Lanka we project expansions in apparel and wood products, and contractions in general manufacturing. All shifts are less than 1%.

Overall, the output shifts tend to indicate that the reductions in international transportation margins would not have a major impact on production structures, and that adjustment costs would be minimal. As expected, the production shifts are more pronounced in the smaller economies in the region, and these areas may need some adjustment assistance, albeit probably minimal.

Download this Paper [ PDF 210.1KB| 26 pages ].

Comment(s)

There are [0] comment(s) for this entry. Post a comment.