A RECURSIVE DYNAMIC CGE ASSESSMENT OF THE CAMBODIAN MILLENNIUM POVERTY REDUCTION TARGET. Sothea Oum. Centre of Policy Studies Monash University

Transcription

1 A RECURSIVE DYNAMIC CGE ASSESSMENT OF THE CAMBODIAN MILLENNIUM POVERTY REDUCTION TARGET Sothea Oum Centre of Policy Studies Monash University Ph: Fax: sothea.oum@buseco.monash.edu.au

2 Preface Title of Thesis: A Recursive Dynamic Computable General Equilibrium Model for Poverty Analyses in Cambodia Supervisor: Professor Philip Adams Summary of the thesis The contribution of the thesis is the first estimated input output table and the building of a major computable general equilibrium (CGE) model for Cambodia. The CGE model features an income distribution and poverty module, which is fully incorporated into the recursive dynamic CGE model. The model can be used for many pressing policy issues in Cambodia. Among the applications illustrated in the thesis are: a forecast simulation of the economy for , an assessment of expected oil windfalls from 2011 onward, and an improvement of the agricultural productivity. The effects on income distribution and poverty are discussed for each simulation. Chapter 1: Introduction and Summary Chapter 2: Theoretical Structure of the Cambodian CGE model Chapter 3: Incorporating Income Distribution and Poverty Framework into the Cambodian CGE Model Chapter 4: Estimation of Input Output Table and Other Required Data Chapter 5: The Application of the Cambodian CGE Model for Policy Analyses Chapter 6: Conclusion and Plan for Future Research The following paper is based on the ongoing research project.

3 Abstract: The main objective of this paper is to apply a recursive dynamic computable general equilibrium model to assess the likelihood of Cambodia meeting her poverty reduction target of her millennium development goals (CMDGs) in Results from the model s forecast simulation imply that Cambodia could potentially reduce its poverty headcount from 35 percent in 2004 to 21 percent in 2015, which is 4 percent below its CMDG s target. However, this optimistic forecast is entirely based on the assumption that the pattern of income distribution throughout the forecast period is the same as in the base period.

4 A RECURSIVE DYNAMIC CGE ASSESSMENT OF THE CAMBODIAN MILLENNIUM POVERTY REDUCTION TARGET 1. INTRODUCTION After years in conflict, Cambodia has re-emerged and become one of the best performing economies in the last ten years. The country s annual economic growth has been 8 9 percent over a decade, due largely to economic reforms to attract foreign direct investment, foreign aids, and to significant increases in garment and tourism exports. In 2004, Cambodia became the first least-developed country admitted to the World Trade Organization (WTO). The country has enjoyed double-digit growth rates since then. According to the International Monetary Fund (IMF), Cambodian economy is expected to grow at about 9 percent in 2007, IMF (2007). Despite high economic growth, the country s achievement in poverty reduction is moderate. Cambodia aims to reduce poverty by half in 2015, according to her millennium development goals (CMDGs). The original CMDG s poverty reduction target was to reduce poverty from an old estimate at 39 percent in 1993/94 to 19.5 percent in However, using backward extrapolation from the socioeconomic survey 2004, the World Bank (WB) has revised the estimation of the poverty headcount in 1993/94 up to 47 percent, WB (2006). The rate in 2004 is estimated to be at 35 percent, and the CMDG s poverty reduction target in 2015 is revised to be at 24 percent. In its assessment, the WB maintains that assuming a 10 percent per annum growth rate in industry and services, and a 2.5 percent growth rate for agriculture would reduce the poverty rate to 29 percent in But the rate would fall to 21 percent, were the annual growth in agriculture 4 percent. In both cases, the growth rate of the economy was assumed to be 7 percent per annum from The purpose of this paper is to employ a recursive dynamic computable general equilibrium model (CGE) to re-assess the above assertion based on economic forecast for the period in question. The rest of paper is organized as follows. In Section 2, we briefly overview the CGE model of Cambodia. We then, in Section 3, describe how to link CGE results with household survey data to conduct poverty analyses. Section 4 discusses poverty implications from the forecast. The conclusion of the paper is in section 5. 1

5 2. THEORETICAL STRUCTURE OF THE CAMBODIAN CGE MODEL The staring point of the core model is ORANI, the Australian static CGE model. The main theoretical features of the model can be found in Horridge (2000) and the detailed description of the model is provided by Dixon et al. (1982). The model consists of equations describing, for some time period: producers' demands for produced inputs and primary factors; producers' supplies of commodities; demands for inputs to capital formation; household demands; export demands; government demands; the relationship of basic values to production costs and to purchasers' prices; market-clearing conditions for commodities and primary factors; and numerous macroeconomic variables and price indices. Demand and supply equations for private-sector agents are derived from the solutions to the optimisation problems (cost minimisation, utility maximisation, etc.) which are assumed to underlie the behaviour of the agents in conventional neoclassical microeconomics. All markets are cleared and the agents are assumed to be price takers, with producers operating in competitive markets which prevent the earning of pure profits. Following Johansen (1960), the model is solved by representing it as a series of linear equations relating percentage changes in model variables using GEMPACK developed by Harrison and Pearson (1996). The model was calibrated with our estimated input output dataset for Cambodia, Sothea (2007). The data represents the economy in All relevant elasticises are taken from the Global Trade Analysis Project (GTAP) due to the lack of data specific to Cambodia for econometrically estimated parameters. 2

6 2.1 Production Structure We assume that producers minimise their input costs given level of output with nested Leontief/constant returns to scale (CES) production. Production is assumed to be separable in order to reduce the dimension of parameter space in the optimisation problem. At the top level, commodity composites, a primary-factor composite and 'other costs' are combined using a Leontief production function. Consequently, they are all demanded in direct proportion to output. Each commodity composite is a CES production function of a domestic good and the imported equivalent, following the Armington (1969) imperfect substitution. The primary-factor composite is a CES aggregate of land, capital and composite labour, of which land and capital stock are assumed to be industry-specific. Composite labour is a CES aggregate of occupational labour types. 2.2 Final Demands The demand for investment goods are derived from two-part cost-minimization. First, the total cost of each imported and domestic commodity is minimized subject to the CES function. At the aggregated level, the total cost of commodity composites is minimized subject to the Leontief production function. No primary factors are used directly as input to capital formation. The household demand is modelled similar to that of the investment demand. The only difference is that commodity composites are derived by a Klein Rubin utility maximization subject to its aggregate budget constraints, leading to the linear expenditure system (LES). The imported and domestic commodities substitute for each other according to a CES aggregation. We simply allow the aggregate demand of each household to response proportionately to their disposal income from wages, capitals, land rentals, and transfers. Government spending is assumed to be exogenously determined. Finally, export demands are modelled as a reverse function of their price in foreign currency and a constant own price elasticity of demand. 3

7 2.3 Regional Extension We follow the top-down regional extension as described in Dixon et al. (1982) by assuming that each industry uses the same technology in each region. The main data required is a matrix showing how industry output is distributed between regions (provinces) and how other final demands are split using data on provincial output, valueadd, and employment. No regional trade in each commodity is needed. The regional industries are divided into national and local industries. The former produce freely tradable commodities and their outputs are assumed to move proportionate to national output, whereas commodities produced by the latter (mainly services) are scarcely traded across regional borders and their outputs move in line with the local demand for the corresponding commodities. Household consumption in each region is tied to regional labour income. The alternative treatment is to relate regional consumption to total factor payments in the region. However, the employment of non-labour factors in the region does not necessary generate income payments in that region. In any case, the movement in labour income and total income are likely to be similar. These assumptions produce local multiplier effects: regional benefits from both expansion in national industries and from increased demands for local commodities. This extension allows us to translate national simulations results into the regional ones such as regional output, employment as well as the regional consumption that can also be used in the poverty analyses. 2.4 Simple Dynamic Features In order to capture inter-temporal changes in main variables in question, additional recursive dynamics are needed to accommodate stock-flow relations in physical capital accumulation and real wage-employment adjustment. There are 3 main mechanisms added into the core model: (i) a stock-flow relation between investment and capital stock, which assumes a 1 year gestation lag; (ii) a positive relation between investment and the rate of 4

8 profit; (iii) a relation between wage growth and employment. The formal mathematical forms of these features are found in Horridge (2002). Annual rates of growth of capital stocks are linked to investment; investment in turn is guided by rates of return. Starting point of each computation represents the economy as it was both at the end of the previous period and at the beginning of the current period. Similarly, the 'updated' data base produced by each computation represents the economy as it will be both at the end of the current period and at the beginning of the next. Changes in variables compare their values at the end of the current period with those at the beginning of the current period. We allow for real wages to adjust to employment levels as follows: If end-of-period employment exceeds some trend level, then real wages will rise. Since employment is modelled as negatively related to real wages, this mechanism causes employment to adjust towards the trend level, which may be thought of as the level of employment corresponding to the natural rate of unemployment (NAIRU) hypothesis. 3. LINKAGE BETWEEN CGE MODEL AND POVERTY ANALYSES The applications of CGE models in poverty analyses are becoming more common. Filho and Horridge (2004), and Savard (2003 & 2005) provide very helpful literature reviews and good discussions on the topic. According to them, the application of CGE in income distribution and poverty analyses can be classified into three main categories depending on how households are integrated into the CGE models. The first approach is a model with a single representative household (RH) through which poverty analysis can be performed by using the variation of income or expenditure of the RH generated by the model with household survey data to conduct ex ante poverty comparison. Even though this approach is easy to implement, its main drawback is it provides no information on the intra-group income distribution. The second approach is the integrated multiple-household model (MH), in which there are a number of representative households. The main advantages of this approach are that it provides richer information on intra-group income distribution changes and prevents pre- 5

9 judgement from aggregating households into categories. However, the data reconciliation and the size of the model can become a constraint. The third approach is the application of the micro-simulation (MS) techniques. This approach provides richer information on household behaviour (consumption and labour supply) for large record units of household survey data. However, the main drawback of this approach is the lack of consistency and the feedback between the CGE model and the micro-simulation model. In this paper, we apply the method in-between the first and second approach. What we mean by that is we integrate 15 household categories into the model according to their geographical location. We also make use of results from previous surveys to impose intragroup income distribution. We use expenditure to conduct poverty analyses. Using the results from our CGE model, we then perform poverty analyses in two ways. In one way, if the mean consumption of a household category increases, the expenditure of each household within the group is simply raised by the same proportion. In another, rather than updating expenditure of every household by the same proportion, we calculate expenditure elasticities of the sub-group households within each household category based on data from previous surveys. We use these elasticites to calculate the changes in expenditure of each sub-group household in response to changes in the household category s mean consumption. We then compare the poverty levels obtaining in the post-simulation case with those in the pre-simulation case using the FGT index of Foster, Greer, and Thorbecke (1984). Given a vector of household incomes (expenditures) y = (y 1, y 2,...,y n ) in increasing order, a predetermined poverty line z > 0. Where g i = z -y, is the income shortfall of the i th household, q = q(y; z) is the number of poor households (having income no greater than z), and n = n(y) is the total number of households, the index is given by the following formula: 1 P(y; α z) = n i=1 q gi z α 6

10 where: whenα = 0, P 0 is commonly known as the poverty headcount index, the percentage of the population with per capita consumption below the poverty line, whenα = 1, P 1 is the poverty gap index which is the average shortfall of income from the poverty line, and whenα = 2, P 2 is the poverty severity index which gives greater weight to those that fall far below the poverty line than those that are closer to it. When the y vector is broken down into subgroup expenditure vectors y (1),,y (m). The index can also be written as: n P y z m ( ) P(y; α z) = j j ( ; ) j=1 n α Therefore, the total index is the weighted sum of the subgroup levels. 4. POVERTY IMPLICATIONS FROM THE ECONOMIC FORECAST In this section, we discuss steps in conducting the forecast over the period based on macroeconomic data published by the WB and IMF, and Cambodia s National Institute of Statistics (NIS). We follow the methodology pioneered by Dixion and Rimmer (2002) for the MONASH model. 4.1 Forecast Closure and the Stylized Model In the forecast simulation, the solution is on annual basis: the base solution for the year-t computation is the solution for year t 1. This means that, for instance, start-of-year capital stock for year t is completely determined by end-of-year capital stock in the base solution. However, end-of-year capital stock of each year is endogenous and determined through changes in real rate of return (ROR) and end-of-the year investment. We illustrate the closure by Figure 1. Gross Domestic Product (GDP) and agricultural land in each industry are exogenous and can be shocked by the forecast values. The primary-factor efficiency is endogenized to capture economic-wide changes in productivity. Employment and real wage are allowed to adjust with the employment trend (NAIRU). 7

11 Figure 1: Causation in the Forecast Closure Real Wage Employment Trend ROR Legend Endogenous Employment Primary-factor efficiency Start-of-year Capital Stock Exogenous End-of-Year Capital Stock End-of-Year Investment Household consumption Aggregate investment Government spending GDP = Trade balance On the expenditure side, real consumption is endogenous. Real aggregate investment and real government spending are exogenous, allowing them to take shocks from the macro forecasts. Foreign currency prices of imports are naturally fixed as Cambodia is a small and open country. The consumer price index (CPI) is used as numeraire is shocked by the forecast value. National population is also allowed to change with official forecast. Other variables in this closure are such as taxes, tariffs, and quantity shift variables are assumed to be fixed. To track the causal relation between these variables, we adopt a commonly used strategy to illustrate the mechanisms of the core model through the sketch model version of MONASH by Dixon and Rimmer (2002). It is presented in Table 1. (1) Y=C + I + G + X M (2) Y = A F (K, L) (3) C + G = APC*Y (4) C / G = Γ (5) M = H (Y, TOT) (6) TOT = J (X, V) Table 1: Equations of the Stylized Model 8

12 (7) I = N (ROR,RROR) (8) K / L = Q (ROR, A, TOT) (9) W = U (K/L, A, TOT) (10) L/ L = Z(W ) Equation (1) is the GDP (Y) identity in constant-price terms. Equation (2) is the economy s production function relating real GDP to inputs of labour (L), and capital (K) and to input-saving technology term (A). In writing (2) and elsewhere in the stylised model we ignore the existence of agricultural land and the presence of distortions due to indirect taxes and subsidies. Equation (3) links total consumption (C+G) to GDP via a given propensity to consume (APC). Equation (4) defines Γ, the ratio of private (C) to public (G) consumption spending. Equation (5) summarises the determination of import volumes (M). In the absence of changes in preferences, import volumes are positively related GDP and the ratio of domestic to imported prices (represented here by TOT, i.e., the price of exports relative to the price of imports). Commodity exports are inversely related to their foreign currency prices via constant elasticity demand functions. This is summarised by equation (6), which relates the terms of trade to the volume of exports (X) (movements along foreign demand schedules) and a shift variable V (movements in foreign demand schedules). Equation (7) links aggregate investment to the rate of return (ROR) and the shift variable (reflecting investors confidence, RROR). With constant return to scale assumption, the marginal product functions are homogeneous of degree 0 and so can be expressed as functions of K/L and A. This accounts for equations (8) relating the profit maximising capital/labour ratio to the rate of return on capital, technological change (A), and the terms of trade (TOT). Similarly, the movement in the real consumer wage (W) can be related to changes in the capital / labour ratio, technology, and the terms of trade in equation (9). Following Giesecke (2004), we deriving (8) & (9) by solving the firm s profit maximization problem: Π = P.Y W L.L W K.K, subject to Y = A f(l,k); where Π is profit, W L is the wage rate, W K is the rental price of capital, and P is the price of output Y. From this problem we have the f.o.c. Π K = P.A.f K W K = 0, or f K = W K /(A.P), or equivalently f K = (W K /P I )(P I /A.P). Noting that f K is a monotonically decreasing function of K/L, that (W K /P I ) is the rate of return (ROR in equation 8) and that P I /P is a negative 9

13 function of the terms of trade (since P I the investment price index includes import prices but excludes export prices, while P the price of domestic output includes export prices but excludes import prices) provides (8). This implies that Q ROR < 0, Q TOT > 0, and N A > 0. By the same token, Π L = P.A.f L W L = 0, or f L = W L /(A.P), or equivalently f L = (W L /P C )(1/A)(P C /P). Noting that f L is a monotonically increasing function of K/L, that (W L /P C ) is the real consumer wage (W in equation 9) and that P C /P is a negative function of the terms of trade (since P C the consumer price index includes import prices but excludes export prices) provides (9). This implies that U K/L > 0, U TOT > 0, and U A > 0. The real wage must also adjust according the NAIRU rule via equation (10) where L is the employment trend. The stylized model can now be used to describe the main features of the base forecast closure. Under this closure, Y, P, I, G, APC, J and V are determined exogenously. Most equations can be readily associated with the determination of a specific endogenous variable. Hence, we might think of (8) as largely determining K and ROR which in turn determines I in each industry. C and G are determined by equations (3) and (4). Equation (5) determines M leaving (1) to determine X. Equation (6) determines TOT. The real consumer wage is determined by (10) and L in (9). Linkages like these will be used to explain the results of the forecast simulation. 4.2 The Economic Forecast and Poverty Implications The Economic Forecast for We use forecasts published by the NIS, the WB and IMF as inputs into the model. These forecasts are for main macroeconomic variables and for structural variables. However, for the current exercise, we use only the information presented in table 2. We select these variables for the following reasons. First of all, the forecast of GDP growth is used to compare with the extent of poverty implication done by the WB (2006) over the same period. Land clearances and land concessions for agricultural purposes are prevalent. We use the trend of annual growth in the past decade of arable land reported by NIS (2006) and assume that it is carried over projected period. The latest projection in population growth by NIS is also used. Even though labour supply is normally linked 10

14 closely with population growth, we adopt a higher rate of growth in employment trend (NAIRU) to reflect underemployment and the trend rate of growth in labour participation in Cambodia. The forecasted changes in consumer price are also used in the forecast. Table 2. Main Macroeconomic Forecast (percentage change) Gross Domestic Product Consumer Price Index Aggregate Investment Government Spending Employment Trend Population Growth Cultivated Land Area Source: NIS, WB, and IMF Taking these extraneous forecasts into the model, we are ready to investigate their impact on other endogenous variables at both macro and some industrial results using causation diagram and the stylized model. The details of forecasts are presented in table 3. 11

15 Table 3. Macroeconomic Results (percentage change) Real GDP Aggregate household consumption Aggregate real investment Aggregate real government demands Export volume Import volume All primary-factor efficiency Aggregate land stock Aggregate employment Aggregate capital stock Capital/labour ratio Trend employment Average real wage Real devaluation Nominal exchange rate GDP deflator Consumer price index Rate of return index Terms of trade

16 Assuming for the moment that wage is sticky and that A are unchanged, it is clear from the stylized model that: with K exogenous (row 10), L (row 9) is effectively determined (via equation 2). However, when the actual employment grows higher than the employment trend the real wage must adjust upward, and vice versa, to reinforce the NAIRU rule via equation (10). The employment moves back to the trend in 2008 and down afterward before picking up toward the trend along with slowing real wage (row 13). With Y, L and K given, so too is A (row 7) as a residual. With K, L, A, and W given, equation (9) determines TOT (row 19) which is consistent with the movements of the balance of trade described below. They together determine ROR, RROR, X, and M via equation (5) (8). Via equation (8), the increase in capital-labour ratio (row 12) must cause the marginal product of capital to fall (that is, ROR must fall). This accounts for the reduction in the rate of return index (row 20). This in turn causes the marginal product of labour to rise, justifying the increase in the real wage. With aggregate investment I (row 3) given, equation (7) implies that investors are more confident in the country s economy, thus willing to supply more capital at a lower required rate of return (RORR), i.e. rightward shifting in the capital supply curve. With Y, G, and APC exogenous, the household real consumption C (row 2) is determined via equation (3). From equation (1), even though with lower increase in G, the higher increases in real absorption C and I relative to GDP are sufficient to cause the balance of trade moving towards deficit in the first two years. This is achieved through an appreciation of the real exchange rate (row 14). The appreciation in the real exchange rate causes import volumes to rise (row 6) in conjunction with growing domestic demand via equation (5). From equation (6), the increase in export volumes (row 5) reaffirms the deterioration in the terms of trade. The real devaluation causes the improvement in the balance of trade from 2007 onward. The sectoral results, presented in table 4, largely follow from the macroeconomic results. With higher GDP growth driven by both domestic demand and international trade, so too is the activity of all sectors to satisfying these demands. 13

17 Table 4. Industrial External Trade Structure and Output Results Share in total country s export Export share of total output Import share in local market Share in total imports % Accumulated changes in output Paddy OtherCrops Livestock Forestry Fishery Mining FoodProds BevTbacco Textiles WearingApp LeatherFtw WoodPrd PaperPrt OilGasPrd ChemRubPlas NonMetlMin FabBasMtlPrd MotorVehicle TrasportEqui ElectronicEq Machinaries OthManuf Electricity Water Construction WholeSales RetailTrades Repairs HotelRestau OthTransport WtrTransport AirTransport PostComm FinanceServ RealEstBus PubAdmin Education Health OtherServ Dwellings

18 The table shows the expansions in all 40 sectors of the economy. However, some sectors gain more than the others due to the underlying input - output linkages of and between sectors, and their sale pattern. For instance, among the top winners are industries They are mostly importing industries and main suppliers of capital goods. Construction sells largely to dwelling and both gain significantly from the strong growth of the economy. Most of textile imports are used by the wearing & apparel industry. They also both stand to gain together. The second most advantageous industries are in the service sectors which are a mixture of labour intensive, domestic and traded-oriented industries. The last group are agricultural sectors except forestry which enjoy a moderate gain due to their labour intensity. Moreover, these agricultural sectors causes slow growths in the other sectors that use outputs from them as their main inputs. Those are foods and beverages, tobacco, and rubber industries Household Consumption and Poverty Implications As briefly discussed above, the household expenditure is modelled as a linear expenditure system (LES) derived from a Klein Rubin utility. The results of the real expenditure of the 15 household categories are presented in table 5. Table 5. Accumulated Changes in Real Household Consumption by Categories (percent) Banteay Mean Chey Bat Dambang Kampong Cham Kampong Chhnang/Pursat Kampong Speu Kampong Thum Kampot Kandal Phnom Penh Prey Veaeng Siem Reab Sihanouk/Kep/Koh Kong Svay Rieng Takaev Other* *Kratie, Mondul Kiri, Preah Vihear, Ratanak Kiri, Strung Treng, Oddar Meanchey, and Pailin 15

19 It can be seen that the very optimistic forecast causes strong growth in households consumption for all categories. The households in the capital Phnom Penh enjoy the largest gains. If these average gains are applied to the base period consumption of every household in each category, it would drastically drive down poverty in any measure. We believe that it is not a plausible scenario. In stead, we derive the household s expenditure by categories and deciles using the elasticity of decile-household consumption in response to changes in the average consumption of the household category they belong to. It is shown in table 6. Table 6. Accumulated Changes in Real Household Consumption by Categories and Deciles (percent) 2010 Provinces/Cities D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 1 Banteay Mean Chey Bat Dambang Kampong Cham Kampong Chhnang/Pursat Kampong Speu Kampong Thum Kampot Kandal Phnom Penh Prey Veaeng Siem Reab Sihanouk/Kep/Koh Kong Svay Rieng Takaev Other* Provinces/Cities D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 1 Banteay Mean Chey Bat Dambang Kampong Cham Kampong Chhnang/Pursat Kampong Speu Kampong Thum Kampot Kandal Phnom Penh Prey Veaeng Siem Reab Sihanouk/Kep/Koh Kong Svay Rieng Takaev Other* *Kratie, Mondul Kiri, Preah Vihear, Ratanak Kiri, Strung Treng, Oddar Meanchey, and Pailin 16

20 Applying these changes to each household s consumption in the deciles of every household category in the base year, we are able to recalculate the poverty indices. The base year household consumption and poverty indices are estimated by James, (2005). We use the same dataset to estimate poverty indices for 2010 and 2015 as presented in table 7. Since the household consumption is in real terms, we do not need to update the poverty line from the base period. Table 7. The FGT Poverty Indices (percent) Base P0 P1 P2 P0 P1 P2 P0 P1 P2 Cambodia Banteay Mean Chey Bat Dambang Kampong Cham Kampong Chhnang/Pursat Kampong Speu Kampong Thum Kampot Kandal Phnom Penh Prey Veaeng Siem Reab Sihanouk/Kep/Koh Kong Svay Rieng Takaev Other* *Kratie, Mondul Kiri, Preah Vihear, Ratanak Kiri, Strung Treng, Oddar Meanchey, and Pailin In the base year, poverty is prevalent in Kampong Speu, Kampong Thum, Siem Reab, and other small provinces. Capital Phnom Penh and its adjunct Kandal province have the lowest rate of poverty in all measures. The average poverty headcount of the country in the base year is 35 percent. With significant increase in households consumption all over the country, the poverty headcount falls below the CMDG s target of 24 percent in The poverty headcount in 2015 is 21 percent, coincidently the same rate as the best scenario of the WB s assumption of a 4 percent per annum growth in agricultural output. Not surprisingly, Phnom Penh and Kandal manage to reduce more than half of their poverty headcount compared with the base period. They are the centre of the country s industry and main economic activities. The initial poverty-stricken provinces maintain the 17

21 same status at the end of the forecast. These results reflect the fact that we impose the same pattern of income distribution as in the base year. We ignore of the possibility of the dynamics movement of households in and out of poverty. A poverty analysis in a real dynamic sense is still a challenging research issue. Even though our poverty results are the same as the WB s second scenario, we depart from its assumptions as follows. Firstly, we adopted a higher GDP growth rate throughout the forecast period, while the World Bank assumes a straight annual 7 percent growth in GDP from Secondly, rather than plainly assuming a sectoral growth composition, we allow the model to project the sectoral results based on the forecast on main macroeconomic data. 5. CONCLUDING REMARKS In this paper, we apply a recursive dynamic CGE model to forecast the Cambodian economy for the period The results from the model s forecast are used to assess the likelihood of the country meeting her poverty reduction target in The model s forecast simulation implies that Cambodia could potentially reduce its poverty headcount from 35 percent in 2004 to 21 percent in 2015, which is 4 percent below her CMDG s target. Our poverty headcount result in 2015 is coincidently the same as the best scenario of the WB s assessment, without imposing strict assumptions on the sectoral growth pattern. However, this optimistic forecast is entirely based on the assumption that the pattern of income distribution throughout the forecast period is the same as in the base period. As discussed above, if we plainly use the consumption results from the 15 household categories to update consumption of every house belong to them, we would end up having a very low rate of poverty headcount. This intra-group distribution problem, perhaps, can be solved by integrating as many as more households into the model or by using the microsimulation approach. These will be agenda items for the forthcoming research. 18

22 REFERENCES Armington, P "A Theory of Demand for Products Distinguished by Place of Production." IMF Staff Paper 16, , D. P. Vincent, P. B. Dixon, B. R. Parmenter, Sutton, J. and Vincent D. P ORANI: A Multisectoral Model of the Australian Economy. North Holland, Amsterdam. Dixon, P.B. and M.T. Rimmer Dynamic General Equilibrium Modelling for Forecasting and Policy. Contributions to Economic Analysis 256, North Holland, Amsterdam. Filho, J.B., and Horridge, J.M Economic integration, Poverty and Regional Inequality in Brazil. CoPS/IMPACT General Working Paper Number G 129, (July), Centre of Policy Studies, Monash University, downloadable from Foster, J., Greer, J., and Thorbecke, E A Class of Decomposable Poverty Measures. Econometrica, 52 (May): Harrison, W.J. and K.R. Pearson Computing solutions for large general equilibrium models using GEMPACK Computational Economics, Vol. 9: Horridge, J.M ORANI G: A General Equilibrium Model of the Australian Economy. CoPS/IMPACT Working Paper Number OP 93, (October), Centre of Policy Studies, Monash University, downloadable from: monash.edu.au /policy/elecpapr/ op-93.htm Giesecke, J The Extent and Consequences of Recent Structural Changes in the Australian Economy, : Results from Historical and Decomposition Simulations with the MONASH. CoPS/IMPACT General Working Paper No. G 151, (December), Centre of Policy Studies, Monash University, downloadable from: /g-151.pdf International Monetary Fund Cambodia: 2007 Article IV Consultation - Staff Report; Staff Supplement; and Public Information Notice on the Executive Board Discussion. Country Report No. 07/290, (August), downloadable from /external/pubs/cat/longres.cfm?sk= Johansen, L A Multisectoral Model of Economic Growth. ( 2rd edition), North Holland,, Amsterdam. Knowles, J.C A New Set of Poverty Estimates for Cambodia, 1993/94 to A Report to the EAS Country Units of the World Bank, Washington D.C, Mimeo. National Institute of Statistics. 2006a. Statistical Year Book Kingdom of Cambodia. 19

23 Savard, L Poverty and Income Distribution in a CGE Household Micro- Simulation Model: Top Down/Bottom Up Approach CIRPREE Working Paper 03-43, (October), downloadable from: CIRPEE/cahierscirpee /2003/files/CIRPEE03-43.pdf Poverty and Inequality Analysis within a CGE Framework: A Comparative Analysis of the Representative Agent and Microsimulation Approaches. Development Policy Review, 23, no.3: Sothea, O An Estimation of Input Output Table for the Cambodian Economy 2003, V6.2 Documentation - I-O Tables: Cambodia, Center for Global Trade Analysis, Documentation, Purdue University. World Bank Cambodia: Halving Poverty by 2015? Poverty Assessment. Report No KH, East Asia and the Pacific Region, (February). 20