Modeling Framework and Databases

The current study aims to provide estimates of the impact of improvements in road transport infrastructure and the accompanying improvements in trade facilitation in the GMS. Economic modeling of transportation infrastructure improvements was undertaken using the Global Trade Analysis Project (GTAP) model. We started with the GTAP version 7 database (Narayanan and Walmsley 2008) and utilized this with a modified version of the standard GTAP model (Hertel et al. 2009). The multi-regional computable general equilibrium model and database used in this study are widely used internationally, fully documented, and publicly available (see GTAP 2010 for detailed information). We augmented this model and database to facilitate improved modeling of the GMS, including impacts on poverty.

3.1 The GTAP Model

The GTAP model draws on a set of economic accounts for each country or region, with interactions between regions and sectors captured within a consistent framework. The model we use is comparative and static and retains many standard features of the GTAP model (Hertel 1997). We modeled the behavior of private individuals, firms, and governments, along with responses to changing resource and market conditions. Consumers maximize welfare, subject to their budget limitations, while firms maximize profits using the limited resources available in the economy. In particular, primary factors of production are combined with intermediate inputs, including imports, to produce final output. Armington elasticities allow differentiation between imports from different countries in the GMS and elsewhere, specifying the extent to which substitution is possible between imports from various sources, as well as between imports and domestic production. When the impact of improved infrastructure and trade facilitation improvement is simulated, prices and quantities of marketed commodities, along with impacts on incomes and GDP, are endogenously determined within the model.⁴ While retaining the simple yet empirically robust assumptions of constant returns to scale and perfect competition, Section 3.3 describes how we modified the model to shed new light on the distributional consequences of cross-border transport infrastructure projects.

3.2 The GTAP Database

We used version 7 of the GTAP database,⁵ covering 113 countries or regions and 57 sectors, with a base year of 2004. This release of the GTAP database includes all of the GMS countries: Cambodia, Lao PDR, Myanmar, Thailand, and Viet Nam. While the PRC is available, Yunnan Province and Guangxi Zhuang Autonomous Region are not separated; therefore, we included the PRC in the modeling analysis. We aggregated the database in a way that maintains coverage of all GMS countries with relatively heavy disaggregation of sectors of key importance to the region and to poverty impacts. Details of the regional and commodity aggregation are in Appendix 1 [ PDF 13.6KB | 1 page ] and Appendix 2 [ PDF 19.1KB | 1 page ].⁶

The level of trade between GMS economies varies a great deal, depending on the countries and commodities under consideration. Appendix 3 shows the value of intra-GMS trade, as estimated in the version 7 GTAP database. The country with the highest level of intra-GMS exports in the database is Myanmar, with over 40% of total exports going to other GMS countries. The Lao PDR sends over 28% of its exports to the GMS, though trade with Thailand dominates for both Myanmar and the Lao PDR. For Thailand and Viet Nam, 12–15% of exports are destined for GMS countries; however, exports to the PRC dominate. Excluding exports to the PRC, intra-GMS exports are only a little over 3% for Thailand and Viet Nam. Exports from Cambodia to other GMS countries, including the PRC, are relatively low at about 5%. As indicated in Appendix 3 [ PDF 48.3KB | 2 pages ], in addition to variation by country, there is substantial variation by industry.

Of particular relevance to the current study are the international transportation margins included in the GTAP model and database. Margins are included in the database for air, water, and other transportation, with the latter including land transportation and therefore of key importance to our study. Appendix 4 [ PDF 33KB | 1 page ] shows the cost of bilateral GMS land transport margins as a proportion of the value of exports, calculated from the GTAP database. Cross-border land transport costs are likely to be relatively significant for poorer economies with less-developed infrastructure. This appears to be reflected to some extent in the database, with cross-border land transport margins within the region appearing most significant for the relatively poor countries of Cambodia, Lao PDR, and Myanmar.⁷

3.3 Analysis of Poverty Impacts

A general equilibrium approach is needed to predict changes in real earnings stemming from infrastructure improvements in the GMS. Transport infrastructure improvements will not have uniform impacts on poverty across the GMS and we use household survey data to augment the GTAP database so that the implications of infrastructure development for poverty may also be considered for Cambodia, Lao PDR, Thailand, and Viet Nam.⁸

3.3.1 Analytical Framework

There are many approaches to estimating the change in poverty headcount due to trade reforms (Winters, McCulloch, and McKay 2004; Hertel and Reimer 2005; Hertel and Winters 2006). The approach here builds on those outlined in Hertel et al. (2009), and begins with a consumer demand system and the associated utility function. The poverty level of utility was identified based on international poverty levels (using the World Bank's US$1 per day and US$2 per day poverty lines).⁹ Our evaluation of poverty changes thus amounts to calculating the percentage of the population below this poverty level of utility.

We used Rimmer and Powell's (1996) AIDADS demand system (An Implicit Direct Additive Demand System) to represent consumption in the neighborhood of the poverty line. AIDADS is particularly useful for poverty analysis as it devotes two thirds of its parameters to consumption behavior in the neighborhood of the poverty line (Cranfield, Preckel, and Hertel 2006). Estimation of this demand system was undertaken using the per capita consumption dataset offered by GTAP version 7, with the demand system estimates then calibrated to reproduce base year per capita demands in each country.¹⁰ From there, per capita income was shocked back to the international poverty line in order to identify the poverty level of utility and to estimate consumption quantities at the poverty line.

A key finding in the work of Hertel et al. (2004) is the importance of stratifying households by primary source of income. For example, farm households in developing countries may rely on the farm enterprise for virtually all of their income. And national poverty tends to be concentrated in agriculture-specialized households in the poorest countries in our sample. In these cases, the poor are more likely to benefit from farm price increases. In other countries, the national poverty headcount is dominated by wage earners who will be more susceptible to food price increases. To delineate the patterns of specialization in earnings, we followed Hertel et al. (2004) in identifying five household groups that rely almost exclusively (95% or more) on one of the following sources of income: agricultural self employment, non-agricultural self-employment, rural wage labor, urban wage labor, or transfer payments. The remaining households are grouped into rural and urban diversified strata, giving seven strata.¹¹

Given our interest in comparing results across countries, we took a simplified approach to poverty analysis, focusing solely on the poverty headcount—at both the US$1 and US$2 per day levels. We did so by employing a survey-based highly disaggregated poverty elasticity-based analysis. In particular, we adopted from Hertel et al. (2009) the following equation for predicting the percentage change in poverty headcount, ^ˆH_r, in each of the GMS countries for which household survey data are available:

The term in parentheses on the right side of (1) reports the change in the after tax wage rate for endowment j in region r, ^ˆW_rj, relative to the change in the cost of living at the poverty line, ^ˆC^p_r. This real earnings term is pre-multiplied by several important parameters, which deserve additional discussion. The first of these is the share of earnings type j in total income of households in the neighborhood of the poverty line in stratum s of region r,α^p_rsj. This translates a change in, for example, the wage of unskilled labor, into a change in total household income. If wages rise by 10% and this is 95% of household income for households in the neighborhood of the poverty line in the rural wage labor stratum, then income is predicted to rise by 0.95 * 10% = 9.5%. By definition, the earnings shares sum to one, i.e., Σ_jα_rsj = 1, and summing over the share-weighted change in factor returns yields the total income change for households in the neighborhood of the poverty line for a given stratum or region combination.

This change in income is, in turn, multiplied by the estimated elasticity of the stratum-specific poverty headcount, H_rs, with respect to income, ε_rs. In order to turn these stratum changes into the estimated percentage change in national poverty headcount, they must be weighted by each stratum's share in national poverty:

Summing across strata, we obtain the national poverty headcount reported in (1).

3.3.2 Integrating Poverty into the Model

To model poverty impacts within the GTAP framework, we introduced factor market segmentation, which is important in countries where the rural sector remains a dominant source of poverty (Keeney and Hertel 2005). Here, farm/nonfarm mobility is restricted by specifying a constant elasticity of transformation function that limits the mobility of labor and capital between the farm and nonfarm sectors. Therefore, farm and nonfarm factor returns may diverge, and this becomes a key driver for our distributional analysis. In order to parameterize these constant elasticity of transformation factor mobility functions, we drew on the OECD's (2001) survey of agricultural factor markets. We assumed a constant aggregate level of land, labor, and capital employment, reflecting the belief that the aggregate supply of factors is not overly affected by these transport projects, especially in the medium run.

Implementation of (1) required us to map factor earnings in the general equilibrium model to household income sources. Agricultural labor and capital received the corresponding farm factor returns from the general equilibrium model, as did non-agricultural labor and capital. Wage labor for diversified households reported in the surveys presented a problem because information was lacking to allocate it between agricultural and non-agricultural activities. We simply assigned to it the composite wage for labor determined by the constant elasticity of transformation endowment function. Finally, transfer payments were indexed by the growth rate in net national income.

3.3.3 Data and Elasticities

While conceptually simple, this approach to poverty analysis is actually quite data intensive. The household surveys for the Lao PDR and Cambodia were processed (Komoto 2009), and this, coupled with previously processed estimates for Viet Nam and Thailand (Hertel et al. 2004), permitted this approach to be implemented for these four countries in the GMS.

Table 3 [ PDF 37.3KB | 2 pages ] reports the estimated earnings shares in the neighborhood of the US$1 per day poverty line in the four countries, α^p_rsj. Endowments are disaggregated into ten categories: agricultural land, self-employed agricultural labor (both unskilled and skilled), self-employed non-agricultural labor (both unskilled and skilled), wage labor (both unskilled and skilled), agricultural capital, non-agricultural capital, and transfer payments. The most difficult part of estimating these earnings shares derives from the need to impute returns to factors of production when the source of income is self-employment. This is achieved by matching self-employed household members with similar wage-earning individuals in the household survey and assigning the average earned wage for this class of workers (ideally, same sex and age, same skill level, same sector, same region). The residual earnings are assigned to capital in the case of non-agricultural income and shared between capital and land in the case of farming.¹²

From Table 3, we can see a number of important points. In the case of the agricultural stratum, in which households earn more than 95% of their income from agricultural self-employment, the bulk of their income is imputed to unskilled labor income. The poor are poor, in part because they do not control a lot of productive assets. Returns to land and capital are most important in Cambodia, with very little residual remaining after wage imputation in Viet Nam. Non-agricultural, self-employed households (column 2) in the neighborhood of the poverty line appear to get more of their income from non-labor income. This is particularly true in Viet Nam, where this figure reaches 40%.

Turning to the wage labor households (columns 3 and 4), we see that the share of income coming from skilled labor is relatively high for Cambodia and the Lao PDR, the poorest of the four economies. This is perhaps not surprising, as increased education and training is often required in order to access the formal labor market. The rural and urban diversified households are just that—highly diversified. This diversification is further accentuated by the fact that we have created this earnings profile by taking all households within plus or minus 5% (i.e., 10% of the total stratum) of the poverty line in each stratum. This diversified group earns income from agricultural activities as well as nonfarm activities, receives transfer payments (quite significant in the case of Thailand), and receives income from capital (particularly in the case of Viet Nam).

As we have seen from equation (1), the earnings shares translate wage changes into income changes, but it is the poverty elasticities, ε_rs, that translate the latter into poverty changes, by stratum. Table 4 [ PDF 26.3KB | 1 page ] reports these stratum-specific poverty elasticities for the four countries. These are arc elasticities, obtained by examining the change in income as we move across the stratum decile surrounding the poverty line. We expect these elasticities to diminish as the total poverty headcount in the stratum rises (i.e., it is harder to reduce poverty by 1% when it represents 28% of the population, as in the Lao PDR, as opposed to less than 2% in Thailand). Accordingly, in Cambodia and the Lao PDR, the poverty elasticities are under 1.0 in all cases, while it is above 2.0 for all strata in Thailand (US$1 per day poverty line) and is nearly 9.0 in the rural labor stratum of Viet Nam, where there are many households clustered around the poverty line and the poverty headcount is relatively low (see below). For the same reason, the poverty elasticity tends to diminish as we move from US$1 per day to US$2 per day—there are simply more households below the poverty line.

Table 5 [ PDF 20.5KB | 1 page ] reports the share of national poverty in each of these strata, β_rs. This table shows that poverty is predominantly a rural phenomenon, with the bulk of the poor concentrated either in the agricultural (Lao PDR) or the rural diversified stratum (Thailand) or both (Cambodia and Viet Nam).

Table 6 [ PDF 22.7KB | 1 page ] reports the average expenditure share, at producer prices, on the 10 broad commodity aggregates in our consumer demand system, at the two poverty lines (US$1 per day and US$2 per day). Food clearly dominates the budgets of the poorest (US$1 per day) households in all four countries—but particularly for Cambodia and the Lao PDR. Therefore, the cost of living at the US$1 per day poverty line will be very sensitive to the price of foodstuffs. By the time income rises to the US$2 per day level, the share of crop products in total expenditure has declined by nearly half in Cambodia and Viet Nam, while at the same time, the share of total expenditure devoted to manufactures and services, including housing and education, has risen sharply.

With this information in hand, we can evaluate the impact of land transport infrastructure projects in the GMS on poverty.

Download this Paper [ PDF 606KB| 39 pages ].

Comment(s)

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