1

AGRICULTURAL TRADE LIBERALIZATION, PRODUCTIVITY

GROWTH AND POVERTY ALLEVIATION: A DYNAMIC GENERAL

EQUILIBRIUM ANALYSIS

MPIA PROJECT

by

Nadia Belhaj Hassine

&

Zied Ben Salem and Hanene Ouertani

29 October 2008

I. Abstract Computable General Equilibrium (CGE) models have gained continuously in popularity as an

empirical tool for assessing the impact of trade liberalization on growth, poverty and equity.

In recent years, there have been attempts to extend the scope of CGE trade models to the

analysis of the interaction of agricultural growth, poverty and income distribution.

Conventional models ignore however the channels linking technical change in agriculture,

product differentiation, trade openness and poverty. This study seeks to incorporate

econometric evidence of these linkages into a dynamic sequential CGE model including

product differentiation, to estimate the impact of alternative trade liberalization scenarios on

welfare, poverty and equity.

The analysis uses the concept of the metafrontier function in investigating the influence of

trading on agricultural technological change, productivity growth and product varieties. The

estimated productivity gains induced from higher levels of trade are combined with a general

equilibrium analysis of trade liberalization to evaluate the direct welfare benefits of poor

farmers and the indirect income and prices outcomes. These effects are then used to infer the

impact on poverty using the traditional top-down approach and the Tunisian household

survey.

The model is applied to Tunisian data using the social accounting matrix of 2001 and the

2000 household expenditures surveys.

2

II. Main research questions and policy relevance

Agriculture is an economically and socially important sector in Tunisia and remains among

the most distorted sectors due to the heavy use of trade barriers and support policies.

Historically, attempts by the Tunisian government to achieve food self-sufficiency have led to

the implementation of important development projects and regulation measures of the

agricultural and rural activities. The development policy targeted the modernization of the

farming sector, the intensification of the production and the promotion of strategic

commodities. The regulating mechanisms were notably aimed at ensuring adequate income

levels for farmers by reducing their exposure to the food price instability in the world

markets, as well as at preventing consumers from the risk of scarcity in staple commodities.

The government interventions were mainly channeled via the control of prices and the

protection of the domestic market by tariffs and non-tariff barriers.

Faced with structural economic difficulties and mounting financial and sectoral imbalances, it

became evident that large budgetary outlays for the agricultural sector and urban consumers

were losing support on efficiency and affordability grounds. Consequently, the government

started the reform of agricultural policy, which culminated with the adoption and

implementation of the Agricultural Structural Adjustment Programme (ASAP) in 1986 that

aimed particularly at shifting prices closer to those in world markets and reducing production

subsidies.

The economic reform strategy was accompanied by a gradual liberalization of the overall

economy and the promotion of private-sector initiative. The broad trend towards a deeper

integration into the free trade and open market-based world economy was accelerated after the

signature of the General Agreement on Tariffs and Trade (GATT) and joining the World

Trade Organization in the early 1990s. The signing of the partnership agreement with the

European Union in mid-1995, stands for an important step for intensifying the Tunisian’s

economic and financial relations with Europe. While currently limited to the removal of tariff

and non-tariff barriers on manufactured goods, the agreement called for a gradual agricultural

liberalization. A comprehensive free trade in the agriculture sphere is not envisaged at the

present time, the agreement aims simply at consolidating, and in some cases improving, the

existing preferential mutual access for specific agricultural products. Freeing agricultural

trade is however at centre stage of the current Doha Development Agenda negotiations and

Tunisia is actively participating in the actual negotiations.

3

Despite these positive developments in terms of market liberalization and reduced State

intervention in the Tunisian economy, the effective protection remains high in the farming

sector, dominated by unskilled wage-workers and family farmers that are overly dependent on

trade protection and government support. The openness process is likely to have a profound

impact on the agricultural development and poverty in Tunisia as it is expected to induce

significant effects on price, income and incentives for investment and adoption of new

technology.

In a country with limited natural resources, trade openness offers great opportunities through

technical cooperation, quality enhancing and export promotion. A less optimistic view cannot

however deny the challenges facing the most vulnerable rural populations for which rain-fed

farming is the essential livelihood source and that may suffer adverse social and economic

consequences from the growing competitive pressures, increases in agricultural and food

prices and tariff preferences erosion.

As a consequence, the sectoral changes that further markets openness is likely to produce and

their precise timing are difficult to predict. A large number of recent empirical studies have

attempted to estimate the size of the trade liberalization benefits for Tunisia (ESCWA, 2001;

Boughzala et al., 2005; Chemingui and Thabet, 2006; Bibi and Chatti, 2006). These studies

follow a general equilibrium approach and tend to conclude in favour of a positive

relationship between liberalization and poverty reduction. There seems however an important

variation in the size of the simulated effects and the gains are estimated to be relatively small.

This may be explained by the fact that these studies mostly focus on the prices and income

linkages and, even if they broadly confirm the existence of dynamic productivity gains, they

rarely considered the long run growth and employment effects that should be derived from the

factors reallocation, the creation of greater productive capacity and from productivity

improvements.

This study tempts to clarify and to estimate the short and long run effects of alternative trade

liberalization scenarios on agricultural and economic growth in Tunisia and to synthesize

poverty and inequality implications. The methodology used here is based on two links, one

connecting trade openness to farming performance, in the presence of product differentiation,

and another connecting agricultural productivity to economic growth, poverty and equity. The

study incorporates econometric evidence of the productivity linkages into a dynamic

sequential general equilibrium model to capture the additional poverty alleviation that could

be expected from the ongoing dynamic growth effects induced by higher levels of trade.

4

The key questions addressed herein are:

1. What are the effects of trade openness on agricultural growth?

The purpose here is to investigate the key parameters that can serve as a basis for estimating

dynamic agricultural productivity and efficiency gains from increased trade.

The empirical literature has been relatively successful in addressing the issues of trade

liberalization on country’s or industry’s growth, it has yet much less to say about the

association between openness and agricultural performance. The difficulty of empirically

investigating this issue arises from at least three sources.

The first complication stems from the difficulty of obtaining a precise measure of agricultural

trade barriers.

Agricultural commodities are currently protected with a complex system of ad-valorem

tariffs, specific tariffs, tariff quotas, and are subject to preferential agreements. The

determination of the appropriate level of protection is a fairly complex task. The MacMaps

database constructed by the CEPII provides ad-valorem tariffs, and estimates of ad-valorem

equivalent of applied agricultural protection, taking into account trade arrangements (Bouët et

al. 2004). Our data on agricultural trade barriers are drawn from this database1.

The second problem is related to the difficulty of obtaining accurate measures of productivity

and factor inputs.

The prevailing methodology in the literature is based on the growth accounting approach. In

this simple framework producers are assumed to be efficient and to use the same underlying

production technology, total factor productivity estimates are retrieved as a residual from a

production function. These assumptions are questionable, and may lead to biased productivity

estimates (Koetter et al., 2007).

In this study, we avoid these restraining hypotheses and tempt to estimate agricultural total

factor productivity using the latent class stochastic frontier model, which explicitly accounts

for inefficiency. This approach enables to control for farmers heterogeneity through the

simultaneous estimation of the probability of class membership and a mixture of several

technologies (Orea and Kumbhakar, 2004 ; Green, 2005).

The third problem concerns the complexity of disentangling the direction of causality, as there

is a growing prevalence of opinions suggesting that a more plausible explanation for the 1 Available at http://wits.worldbank.org/witsweb/default.aspx.

5

positive association between trade openness and higher productivity is merely the self-

selection of the most productive firms into the foreign market (Clerides et al., 1998; Bigsten

et al., 2004; Biesebroeck, 2005). For a positive correlation between openness and productivity

to constitute evidence of learning from exporting, it is necessary to employ a methodology

which rules out the alternative sources of this correlation. We address the difficulty by using

instrumental variables estimation.

2. What are the linkages and pathways through which trade contributes to poverty alleviation?

Trade–poverty linkages are complex and diverse. From the important pathways through

which liberalization affects the poor, we can distinguish price changes as well as wages and

employment variations, especially those of unskilled workers.

Trade openness is expected to induce important variations in agricultural prices and then in

poor households real incomes, since agriculture represents their main livelihoods source and

their main consumption expenditure.

The degree of transmission of the prices changes to poor households depends however on a

number of specific factors such as the quality of infrastructure, the behavior of markets…, and

their ability to benefit from the new trade environment. It depends on their capacity to adjust

to these prices variations by increasing supplies of products whose price has raised, while

reducing consumption of these same goods.

That is we will ask how agricultural liberalization affects agricultural prices and how

much of any price change gets passed through to the poor.

The price changes are closely interrelated with wages and employment changes. For the self-

employed, prices variations directly affect their income, while for workers commodity prices

need to be translated into factor prices, or employment opportunities before they have an

effect.

This important mechanism operates through the labor markets which play, in medium term, a

strong role in determining the poverty impacts of trade reform. The structure and the

functioning of the labor markets are critical to how trade effects get transmitted to wages and

employment especially those of unskilled workers.

If reform boosts the demand for labor-intensive products, it boosts the demand for labor, and

either wages or employment (or both) will increase. However, if the poor are mostly

6

unskilled, while it is semi-skilled labor that receives the boost, poverty will be unaffected. It is

also important to check where the various wage rates lie relative to the poverty line and

whether wages cross critical thresholds to evaluate the poverty impacts.

Assessing the impact of alternative scenarios of agricultural trade liberalization on

factor prices and investigating the functioning of the labor market in Tunisia are also

important questions that we will try to address.

An alternative route through which trade liberalization operate on poverty is productivity

growth. Most of the evidence suggests that a sustained poverty reduction, in the long run,

hinges critically on economic growth that is mainly triggered by improved productivity.

Exposure to international trade stimulates technical change through the diffusion of new

technologies especially from more advanced countries at the technological frontier to

developing countries. Increased imports also provide increased competitive pressures that

help prompt domestic firms to improve their technologies.

Since area expansion and irrigation have already become a minimal source of output raise in

Tunisia, agricultural growth will depend more and more on yield-increasing technological

change. The county’s capacity to benefit from opportunities arising from the new trade

environment is closely related to the producers’ ability to adopt advanced agricultural

technologies. This can be a powerful force in boosting farming productivity growth and in

fostering rural growth which can be inherently pro-poor. Trade liberalization may however

be accompanied by skill-biased technical change, since skilled labor may benefit relative to

unskilled labor if traded sectors draw mainly on the first.

This analysis attempts to examine the effects of trade openness on agricultural

productivity and to asses how farming performance impinges on inequality and

poverty.

The model incorporates product differentiation, as with the fall of tariff barriers, quality

assurance is likely to become the major constraint for international market access. The

product differentiation assumption help to expand the responsiveness of trade patterns to

small price shocks, and allows to account for the fact that Tunisia is both an exporter and an

importer of a given commodity ( for example it is an exporter of high-quality wheat, while

importing low quality wheat) (Winters, 2005).

7

The study aims finally at providing guidance for the necessary complementary policies

that can make trade policy reform more effective in poverty alleviation.

III. Scientific contribution of the research

Plethoras of AGE analysis have focused on assessing the expected benefits of trade

liberalization on poverty. The complexity of the trade-poverty linkages has led analysts to

abstract from some pathways through which trade reforms affect the poor in order to keep the

models tractable. Much of the research on poverty has particularly focused on the

consumption side of the trade-poverty linkages and neglected the factor markets as well as the

long-run productivity mechanisms.

The analyses that addressed factor market linkages between trade and poverty are mainly

premised on the assumption of the eventual translation of the commodity prices changes into

factor markets changes which might entail rural wages variations and therefore affect

household welfare. Ravallion (1990) addressed this issue in a partial equilibrium model that

seeks to measure both the short- and long-run impacts of an increase in the price of rice on

rural wages and poverty in Bangladesh. Two studies by Porto (2003a, 2003b) offer a

generalization of Ravallion’s work. These studies used a general equilibrium approach to

estimate the impact on wages of potential changes in domestic commodity prices arising from

trade reforms in Argentina. The same approach was also used by Nicita (2004) to evaluate the

effects of Mexican trade liberalization on wages.

These analyses ignored however the additional dynamic productivity effects induced by trade

on product and factor prices.

A general equilibrium analysis of technical change in the Philippines by Coxhead and Warr

(1995) revealed important earnings effects resulting from the increase of agricultural

productivity. De Janvry and Sadoulet (2001) explored the implications of agricultural

technology adoption on world poverty and found that price and income effects of agricultural

productivity growth are important in reducing poverty. While these analyses underscored the

critical role of factor markets when examining the poverty impacts of external shocks, these

are not a trade liberalization studies.

8

Only some multi – country trade models tempted to model the links among trade

liberalization, productivity growth and poverty. Cline (2004) included econometrically

estimated productivity gains from increased trade in a CGE analysis of the global poverty

implications of trade liberalization. Anderson et al. (2005) also considered the productivity

effects in the World Bank LINKAGE model. While reported in the same publications as CGE

model results, these productivity effects are off-line calculations and, as seen with Cline, the

analyst often reviews the available literature on productivity and trade, deriving a simple ratio

or expected effect.

The off-line productivity calculations need a careful review of the literature which takes to

follow a long and arduous path. The response of productivity to trade liberalization is a

subject of a highly controversial debate among the economists. The estimated productivity

gains from trade diverge as well broadly across studies and countries, which suggest some

uncertainty about the magnitude of the productivity gains. Our study tempts to extend the previous analyses in two directions. First, it seeks to

investigate the key parameters that can serve as a basis for estimating dynamic agricultural

productivity gains from increased trade. Second it incorporates econometric evidence of the

productivity linkages into a dynamic sequential AGE model with product differentiation, to

arrive at a comprehensive calculation of alternative trade liberalization scenarios on

commodity and factor prices, as a basis for then calculating the corresponding impact on

poverty.

IV. Methodology The methodology is based on two distinct approaches:

- An econometric approach in order to:

1. Estimate some elasticity values to complete the calibration of the CGE model.

2. Investigate the relationship between trading, agricultural productivity and

product quality.

- A CGE analysis of the macroeconomic implications of the agricultural trade openness.

IV.1 The Econometric model

We begin with a partial equilibrium approach to estimate the model’s elasticities as allowed

by the data availability and to assess the contribution of trade openness to the Tunisian

9

agricultural performance. The stochastic estimation of the elasticities values requires

empirical measures of product quality and agricultural total factor productivity. Product

quality estimates are inferred from market shares and unit-value trade data using the discrete

choice framework proposed by Berry (1994). Agricultural TFP measures are obtained using

the latent class stochastic frontier modeling approach (Orea and Kumbhakar, 2004; Greene,

2005)

IV.1.1 Quality Measurement

Our procedure to infer quality estimates relies on a discrete choice representation of product

differentiation. According to this approach product quality is defined as all observable

characteristics plus an intangible attribute that influences the consumer’s valuation of the

good. Given the vector of prices for all available goods, products with larger market shares

are considered as having higher quality. The model allows for heterogeneous preferences

through the interaction of consumer and product characteristics in a nested logit demand

system (Berry, 1994; Kraay et al., 2002). Prior to the estimation, the available varieties,

indexed by j: 1,…,J, are grouped in several product categories, denoted g: 1…G. Varieties

within the same category are considered as closer substitutes than those belonging to different

categories.

We consider I trading countries, indexed by i: 1..I, producing and exporting the variety j in

category g. Consumers have heterogeneous tastes indexed by n. Each chooses a single unit of

the variety that gives him the highest utility. As is standard in logit specifications, we consider

an outside option (g = 0) where the consumer does not purchase any of the available varieties.

Ignoring time subscripts, the indirect utility of consumer n from buying variety j∈ g is:

nijnigijniju εζδ ++= (1)

ijijij pαξδ −= (2)

0n0n00nu εζξ ++= (3)

the utility specification in (1) is composed into a mean utility component, denoted by ijδ and

two unobserved error components that capture individual taste differences. nigζ represents

consumer n’s idiosyncratic taste for products in group g, which varies only across the nests.

nijε is consumer’s idiosyncratic taste for good j, which varies only within product sets.

10

( )εζ + andε are assumed to have type-I extreme value distribution. The mean utility

component in (2) contains ijp the variety’s price and ijξ the unobservable component of

utility that can be understood as the consumer’s valuation of product characteristics and then

be used as a proxy for quality. α is an unknown parameter to be estimated. Equation (3)

represents the utility of the outside option which is normalized to 0.

Each agent chooses the product that confers him the highest utility. Integrating over

consumers yields the standard nested logit expression for the conditioned market share of the

product ij within group g:

( )

( )∑ −

−=

k,h hk

ijg/ij )1/exp(

)1/exp(s

σδ

σδ gk,j ∈ and i,h: 1…I (4)

where σ is a substitution parameter among, versus within, the groups.

Similarly, the demand for group g varieties as a share of total demand is:

∑ −

−

+=

g,h1hg

1ig

igD1

Ds

σ

σ g:,1,….,G (5)

where ( )∑∈

−=gk

ikig )1/exp(D σδ , and the outside good share is:

∑ −+

=g,h

1hg

0D1

1sσ

(6)

the demand of the variety ij as a fraction of total demand is:

igg/ijij sss = (7)

Combining expressions (4) to (7) and taking logs gives the following linear estimating

equation:

ijg/ijij0ij )s(Logp)s(Log)s(Log ξσα ++−=− (8)

We expect unobserved product characteristics to be correlated with varieties prices since

equilibrium prices are determined by observed and unobserved product attributes. Within

market shares are also expected to be correlated with ijξ . If we consider ijξ as an error term

we can estimate α and σ by instrumental variables. We use the parameter estimates to back

out product quality from (8).

11

IV.1.2. Agricultural Productivity Measurement: Panel Data Specification of a Latent

Class Stochastic Frontier Model

Among the several alternative conceptual approaches to estimating agricultural efficiency and

multifactor productivity, stochastic frontier models have become very popular. Based on the

econometric estimation of the production frontier, the efficiency of each producer is measured

as the deviation from maximum potential output. Evenly productivity change is computed by

aggregating over technology change, factor accumulation, and changes in efficiency.

According to the frontier approach all producers use a common underlying production

technology. However, agricultural producers that operate under various production and

environmental conditions might not share the same production possibilities. Ignoring the

technological differences in the stochastic frontier model may result in biased efficiency

estimates as unmeasured heterogeneity might be confounded with producer-specific

inefficiency (Orea and Kumbhakar, 2004).

The recently proposed latent class stochastic frontier model (LCSFM) has been suggested as

suitable for modeling technological heterogeneity (Greene, 2005). This approach combines

the stochastic frontier model with a latent sorting of individuals into discrete groups.

Heterogeneity across producers is accommodated through the simultaneous estimation of the

probability for class membership and a mixture of several technologies.

Following Greene (2005), we assume the simultaneous coexistence of J different production

technologies. Producers in the sample are grouped into different clusters, each using one of

these technologies. The number of groups and the class membership are a priori unknown to

the analyst. The technology for the jth

group is specified as:

ititjitit u),x(fln)yln( −+= νβ (9)

subscript i indexes producers (or countries) (i: 1…N), t (t: 1…Ti) indicates time and j (j: 1, …,

J) represents the different groups. βj is the vector of parameters for group j, yit and xit are,

respectively, the production level and the vector of inputs. For each group, the stochastic

nature of the frontier is modeled by adding a two-sided random error term vit, which is

assumed to be independent of a non-negative inefficiency component uit.

In order to estimate (9) by the maximum likelihood method we assume the noise term vit to

follow a normal distribution ),0(N 2jνσ and the inefficiency term ui to be a non-negative

truncation of a normal random variable:

12

( ) itjitit ,zgu ωδ= (10)

where itz is a vector of country’s specific control variables associated with inefficiencies

including: land distribution, land quality, water resources, land fragmentation, climatic

variables…., jδ a vector of parameters to be estimated, and itω a random variable with a

half normal distribution. The scaling function ( )δ,zg is modeled as:

( ) ( )( )jitjit 'zlnexp,zg δδ =

In a latent class model, the unconditional likelihood for country i is obtained as a weighted

average of its j-class likelihood functions, with the probabilities of class membership used as

the weights:

∑=J

1:jijiji PLFLF (11)

where LFi and LFij are respectively the unconditional and conditional likelihood functions for

country i, and ijP the prior probability assigned by the researcher on this country to belong

to class j. The salient feature of the latent class model is that the class membership is

unknown to the analyst, the probabilities in this formulation reflect the uncertainty that the

researchers might have about the true partitioning in the sample. To constrain these

probabilities to sum to unity, we parameterize ijP as a multinomial logit model:

∑

=

jij

ijij )q'exp(

)q'exp(P

λλ

(12)

where qi is a vector of country’s specific, but time-invariant, variables explaining the

probabilities and λj are the associated parameters.

Various algorithms for the maximum likelihood estimation have been proposed. The

conventional gradient methods and the expectation maximization (EM) algorithm are among

the most used approaches (Greene, 2001; Caudill, 2003; Orea and Kumbhakar, 2004). Using

the parameters estimates and Bayes' theorem, we compute the conditional posterior class

probabilities from:

13

∑

=

jijij

ijijj PLF

PLFP i| (13)

Every country is assigned a specific class according to the highest posterior probability i.e.,

country i is classified into group k (: 1,…, J) if i|i| jj

k PmaxP = . Furthermore, the estimated

posterior probabilities help to compute the efficiency scores. Given that there are J groups,

the latent class model estimates J different frontiers from which the inefficiencies of the

producers can be computed by two methods. The first method estimate technical efficiency

using the most likely frontier (the one with the highest posterior probability) as a reference

technology. This approach results in a somewhat arbitrary selection of the reference frontier

that can be avoided by evaluating the weighted average efficiency score as follows:

j|i|it itJ

1:jj TElnPTEln ∑= (14)

where )uexp( it jjit ||TE −= is the technical efficiency of country i using the technology of

class j as the reference frontier.

Once this model is estimated, it is possible to assess the rate of total factor productivity

change from the results. Productivity growth is composed of technological progress,

efficiency improvement and scale economies. Consequently, TFP growth can be computed

from2:

ScaleTETCTFP ++=••

(15)

where t

flnTC∂

∂= is technical change which measures the rate of outward shift of the best-

practice frontier, t

|uTE jit

∂

∂−=

• is efficiency change over time, and

( ) •

∑−

= kk

jkj

j xScale εε

ε 1

is the scale effect when inputs expand over time. jε is the sum of all the input elasticities3

kjε .

2 See Kumbhakar and Lovell (2000) for the tri-partite decomposition of productivity growth. 3 Since input elasticities vary across groups, productivity change estimates from equation (15) are group-specific. Unconditional productivity measures can be obtained as a weighted sum of these estimates.

14

The expression for TFP change can be extended to incorporate technology transfer as a source

of productivity growth for producers behind the technological frontier. Griffith et al. (2004)

and Cameron et al.(2005) emphasize the importance of international trade and human capital

for enhancing technology transfer and productivity growth. In these models technology gap is

used to capture the potential for technology transfer, and is included as both a level and an

interaction term to capture an effect on the speed of technology transfer. Following these

authors we derive an equation for agricultural productivity growth as:

itititit

itititititiit

X'GAPITGAPHGAPITHGTFP

υμθθθααα

+++++++=

−−−

−−−−−

1113

112111211 (16)

where GTFPit is the growth rate of agricultural TFP of country i at time t, H is the human

capital level of the country, IT is a measure of international trade, GAP is the technology gap,

and X is vector of control variables including institutional factors 4. iα is a country-specific

constant and itυ is an error term.

Technology gap indicates the deviation of country frontiers from the best practice technology

labeled as metafrontier (Battese et al., 2004). We estimate the metafrontier by taking the outer

envelop of the group specific frontiers, ( )jitj

*it ,xfmax),x(f ββ = . Then we measure the

technology gap as the ratio of the output for the frontier production function for group j

relative to the potential output defined by the metafrontier, ( )*it

jitit ,xf

),x(fGAP

β

β= .

IV.2. The General Equilibrium Model

The effects of agricultural trade liberalization on poverty in Tunisia are assessed with a

dynamic sequential general equilibrium approach including imperfect competition and

product differentiation. The basic features of the model are inspired from the prototype model

of Van der Mensbrugghe (2005), Rattsø and Stokke (2005), Diao et al. (2005), and

Chemingui and Thabet (2006).

The model is calibrated to data from a Tunisian social accounting matrix for 2001. It

distinguishes 33 production sectors, including 23 agricultural and food activities with 10 4

•TFP in equation (7) can be considered as the empirical counterpart of GTFP.

15

urban industries and services. Factors of production are classified as capital, land, labor and

natural resources. Land is further differentiated according to the perennial features of the

crops, the irrigation intensity and the varieties grown. Labor is classified by the level of

qualification (skilled and unskilled) and is disaggregated in five components. Institutions

include households, companies, government and foreign trading partners. The household bloc

is disaggregated into rural and urban households. The trading partners are decomposed into

European Union countries and rest of the world.

The following sections provide an overview on the model structure, a more complete

specification is given in the Appendix.

IV.2.1 The model structure

Production structure The model’s production functions are of the nested structure. Perfect complementarity is

assumed between value added and the intermediate consumptions in each sector. Value added

is a Cobb Douglas (CD) function of labor, land, capital and natural resources. Labor is a CES

bundle of skilled and unskilled labor. Land is also decomposed by type in a CES nest. Land is

agriculture specific and labor is assumed to be fully mobile. Capital and natural resources are

assumed to be sector specific. The model incorporates product differentiation by quality in

the agricultural and manufacturing sectors.

Demand structure

In the demand side, the preferences across sectors are represented by the LES (Linear

Expenditure System) function to account for the evolution of the demand structure with the

changes in income level.

The consumption choices within each sector are a nesting of CES functions. The subutility

specifications are an augmented version of the Dixit-Stiglitz structure of preferences designed

to capture the particular status of domestic goods, together with product differentiation

according to geographical origin as well as horizontal and vertical differentiation. Quality

enters as a utility shifter with the horizontal differentiation between varieties.

Total demand is made up of final consumption, intermediate consumption and capital goods.

Sectoral demand of these three compounds follows the same pattern as final consumption.

16

Overview of the model

Ignoring the difference between foreign and domestic goods, we assume that consumer’s

utility depends on consumption of the output of several industries (i), each of which contains

a large number of differentiated varieties (ω) produced by heterogeneous firms. We assume

that the upper tier of utility determining consumption of the different goods is LES and that

the lower tier of utility determining consumption of varieties takes the CES form,

( ) i

iii CCU α∏ −= min (17)

Ci is a consumption index defined over consumption of individual varieties, q i (ω), with dual

price index, P i, defined over prices of varieties, p i (ω),

( )( )ρ

ω

ρ ωωωθ

1

)(⎥⎥⎦

⎤

⎢⎢⎣

⎡= ∫

Ω∈ i

dqC iii , ( )( )

σ

ω

σ

ωωθω −

Ω∈

−

⎥⎥⎦

⎤

⎢⎢⎣

⎡⎟⎟⎠

⎞⎜⎜⎝

⎛= ∫

11

1

i

dp

Pi

ii (18)

with σ

σρ 1−= , σ the elasticity of substitution between any two goods and ( )ωθ the quality

of variety (ω) . Let QI ≡ Ci and PI ≡ Pi the aggregate industry good and price respectively. The optimal consumption and expenditure decisions are given by5:

( )( ) II

iii Q

Pp

qσ

σ ωωθω

−−

⎟⎟⎠

⎞⎜⎜⎝

⎛=

)()( 1 , ( )( ) I

I

iii R

Pp

rσ

σ ωωθω

−−

⎟⎟⎠

⎞⎜⎜⎝

⎛=

11 )(

)( (19)

with ( )ωωω iii rqp =)()( and III QPR = The production side of the model follows Melitz (2003) and Bernard et al. (2006) in that

production involves a fixed and variable cost every period, and only variable costs move

systematically with firm productivity. Production requires multiple factors of production

whose intensity of use varies across industries. We assume that the production function takes

the following Cobb Douglas form:

∏=k

kiikixAy β (20)

with iy the output, iA the total factor productivity (TFP), kix the inputs and kiβ the input

elasticities.

5 See Melitz (2001) and Bernard et al.(2006) for a similar formulation.

17

Following Diao et al.(2005), we introduce labor augmenting technical progress AL, which is

equal across the different sectors, and land augmenting technical progress AD:

Total factor productivity is then expressed as6: Ldi

Li

DLi AAA ββ= (21)

Where Ljβ and Ld

jβ are the shares of labor and land costs respectively.

Productivity dynamics

Productivity growth is generated through technology adoption and own innovations.

Technology adoption is assumed to combine the gap to the technological leader, defining the

learning potential through imitation; human capital, indicating the ability to exploit foreign

technology; and the level of foreign trade which represents the channel transmitting the new

technology to domestic producers. The basic idea behind this model has been developed by

Nelson and Phelps (1966) in the context of a simple neoclassical growth model. Different

modified specifications of this model have been empirically documented by Benhabib and

Spiegel (2003), Stokke (2004) and Rattsø and Stokke (2005). Our framework to analyze

productivity dynamics is inspired by the Rattsø and Stokke (2005) approach, where TFP

growth rate is related to human capital, measured by the share of public expenditure in GDP,

and foreign trade, measured by the share of total trade in total production. This approach is

consistent with the catching-up hypothesis, where the relationship between productivity

growth and technology gap is linear.

( ) ( )TGAPXSGH

GDPGA −⎟

⎠⎞

⎜⎝⎛+⎟

⎠⎞

⎜⎝⎛= 1

32

1 αα

α

λ (22)

where  the proportional change in TFP, G is public expenditure, TRADE total trade, GDP

gross domestic product, XS is aggregate output, and TGAP is the technology gap. λ, θ1 and

θ2 are constant parameters.

Consistent with Rattsø and Stokke (2005), the first term on the right-hand side is the

contribution from innovation activities, while the second term is the technology adoption

function.

6 Equation (21) holds for the agricultural sector, TFP in the manufacturing sector is:

Li

Li AA β= .

18

As increased openness may lead to skill biased productivity growth, we investigate this effect

through the following CES specification of aggregate labor demand. Following Rattsø and

Stokke (2005) aggregate labor demand is specified as:

[ ] lllll SLAULAL LiLiiρρρρρ ββ γγ1

,2,12/12/1 +− += (23)

The direction and degree of technological bias is introduced through the parameter β, which

gives the elasticity of the marginal productivity of skilled relative to unskilled labor with

respect to labor augmenting technical progress. For β equal to zero, technical change is

neutral and does not affect the relative efficiency of the three labor types. With a positive

value of β technical change favours skilled workers, while negative values imply that

improvements in technology are biased towards unskilled labor.

The reduced form specification of technological bias is assumed to be an increasing and

convex function of adoption relative to innovation:

⎟⎟⎟

⎠

⎞

⎜⎜⎜

⎝

⎛

−⎟⎠⎞

⎜⎝⎛=

•

1

2

XSTRADEαβ (24)

where α is a constant parameter.

Land augmenting technical progress is simply related to international trade as follows:

TRADEAD γ= (25)

IV.2.2. Income distribution and poverty

This section discusses incomes distribution and attempt to provide a brief overview on the

methodology used to analyze the external choc effects on poverty.

The common poverty measures can be formally characterized in terms of per capita income

and relative income distribution as follows:

( )( )pL,YPP = (26)

where Y is per capita income and L(p) is the Lorenz curve. P denotes the poverty measure

which we assume to belong to the Foster-Greer-Thorbecke class (1984):

19

( )dyyfz

yzPz θ

θ ∫ ⎟⎠⎞

⎜⎝⎛ −

=0

, where θ is a parameter of inequality aversion, z is the poverty

line, y is income, and f(.) is the density function of income. 210 PP,P and are respectively the

headcount ratio, the poverty gap and the squared poverty gap.

The analysis of the poverty impacts of agricultural trade liberalization and productivity

growth is based on simulations of the model described earlier using the SAM for 2001 as

base. The model calibration is based on the SAM and the econometric results obtained from

the previous section.

The model is designed of such way to capture the direct and indirect effects of agricultural

trade liberalization on commodity and factor prices as a basis for then calculating the

corresponding impact on poverty. The poverty implications of alternatives trade liberalization

scenarios are inferred using the traditional “top-down” approach.

We first simulate the CGE model to generate full vector of commodity and factor prices

owing to policy experiment. These are then fed into a microsimulation framework to conduct

a detailed analysis of income distribution and poverty at the household level using the

Tunisian household survey of 2000. Following Bibi and Chatti (2006), we use the concept of

equivalent income defined as the level of income that would allow achieving the same utility

levels under different budget constraints. Assuming a Stone Geary utility function, the

equivalent income for each household m within the group h and at each period t can be

written as:

( ) ∑∑ +⎟⎟⎠

⎞⎜⎜⎝

⎛−⎟

⎟⎠

⎞⎜⎜⎝

⎛Π=

i

minh,i,i

i

minh,it,i

m,ht

t,i

,i

i

m,htte CpCpy

pp

y,p,pYi,h

00

0

β

(27)

where minh,iC is the minimum level of consumption of commodity i.

In order to better capture the effects of prices and income variations on poverty, we write the

poverty measures in terms of equivalent income as follows: θ

θ ∑Ρ∈

⎟⎟⎠

⎞⎜⎜⎝

⎛ −=

m

h,me

h,m zYz

nN

P 1 (28)

where nm,h is the household size, N is the population size and Ρ is the set of all poor

individuals.

The basic requirement for the measurement of poverty is the definition of a poverty line in

order to delineate the poor from the non-poor. We follow Decaluwé et al.(1999) and Sánchez

Cantillo (2004), by using endogenous poverty lines produced by the CGE model in order to

20

reduce the potential bias in the measurement of the poverty outcomes that results from

neglecting the effects of trade policy on basic consumption.

The poverty line is estimated as ∑=

f

minff Cpz (29)

where minfC and pf are the quantities and prices of the basic needs by commodity.

V. Data requirement and sources Our study requires an important database to conduct the econometric and the CGE analysis:

V.1. The econometric analysis

The econometric application considers panel data at the national level for agricultural

productions in several south Mediterranean countries involved in the partnership agreements

with the EU and different EU Mediterranean countries presenting a strong potential in

agricultural production. The data set include observations on the main crops grown in these

countries, inputs use, determinants of market competition and countries characteristics. The

variables used in the empirical analysis are summarized as follows:

The econometric application is based on panel data at the national level for agricultural and

agro-food production in the main trading partners and competitors of Tunisia with

demonstrated performance in agriculture, namely Algeria, Egypt, Israel, Jordan, Lebanon,

Morocco, Syria, Turkey, France, Greece, Italy, Portugal and Spain during the period 1990-

2005. Our data set includes observations on the main commodities produced in these

countries, inputs use, international trade, human capital, agricultural research effort, land

distribution, land quality, climatic conditions, institutional factors, per capita income, and

income inequality. These variables are grouped in different sets to estimate the discrete choice

model in (8), the stochastic production function in (9); the parametric function of the

inefficiency component in (10) and the productivity change equation in (16); and the class

probabilities in (12). The data are the FAO (FAOSTAT), World Bank (WDI), AOAD,

Eurostat, CEPII, AMAD, ASTI, UN-WIDER, Barro and Lee (2000), Pardey et al. (2006), and

Kaufmann et al. (2007) databases as well as from the different reports of the FEMISE, FAO,

CIHEAM and ESCWA.

The variables used to estimate the discrete choice model consist of 57 agricultural and agri-

food commodities grouped in 10 sets. The stochastic production frontier is estimated using

thirty six agricultural commodities belonging to six categories and five inputs (cropland,

21

irrigation water, fertilizers, labor and machines). The agricultural product categories include

the main produced and traded commodities in the Mediterranean region. Substantial

protection measures exist in the form of entry prices and customs tariffs. These measures aim

at restricting the exchange of commodities considered as a potential source of strong

competition in the Mediterranean basin, and for which greater openness may have serious

domestic economic and social consequences. The data for the input use by crop for each

country are constructed according to the information collected from recently published reports

from the sources above.

The inefficiency effect model and the productivity growth equation incorporate an array of

control variables representing openness to trade, human capital, land holdings, agricultural

research effort, land quality, and institutional quality. Three different measures are used to

proxy the degree of openness of each country, the ratio of agricultural exports plus imports to

agricultural value added, agricultural trade barriers, and the share of agricultural machinery

and equipment imports in agricultural value added. Agricultural commodities are currently

protected with a complex system of ad-valorem tariffs, specific tariffs, tariff quotas, and are

subject to preferential agreements. The determination of the appropriate level of protection is

a fairly complex task. The MacMaps database constructed by the CEPII provides ad-valorem

tariffs, and estimates of ad-valorem equivalent of applied agricultural protection, taking into

account trade arrangements (Bouët et al. 2004). Our data on agricultural trade barriers are

drawn from this database.

Human capital is measured by average years of schooling in the population over age 25 and is

included to capture the labor quality and the ability to absorb advanced technology. The

inefficiency model includes land quality, which is measured by the percent of land under

irrigation; land fragmentation, which is controlled for by the percent of holdings under five

hectares; and inequality in operational holdings measured by the land Gini coefficient to

capture these effects. Agricultural research effort is measured by public and private R&D

expenditures. Institutional quality includes various institutional variables considered as

indicators of a country’s governance, namely, political stability, government effectiveness,

and control of corruption. Regarding the determinants of the latent class probabilities, we

consider country averages of five separating variables related to natural and modern input

endowments as well as to climatic conditions. The variables included in the class probabilities

are total number of wheel and crawler tractors, total applied fertilizers, total agricultural land,

average farm size, and rainfall levels.

22

2- The CGE analysis

The calibration of the base-year solution of our CGE model requires a consistent data set,

reflecting the structure of the Tunisian economy. As existing SAMs for Tunisia are unlikely

to adequately reflect the structural features of the national agricultural sector, we compiled a

new SAM for the year 2001. Building a completely new SAM requires however gathering a

huge amount of data; we use a top-down approach to carry out the compilation of the new

SAM. Our procedure follows two main steps. First, we construct a Macro SAM from national

accounts. Second, we disaggregate the Macro SAM by activity and commodity to generate a

Micro SAM. The disaggregation mainly relates to agriculture and agri-food processing

commodities and is implemented using the national-accounts tables and different

complementary sources such as the surveys conducted by the National Institute of Statistics

(INS), the different reports of the Ministry of Finance and Planning, and the Ministry of

agriculture7. This step is carried out in order to match with the commodity structure of the

Tunisian household expenditures, and in a way that is consistent with the national accounts

and coefficients from a prior SAM. As the data discrepancies in the micro matrix may cause

unbalances, we apply the cross-entropy approach to generate a balanced SAM table. Table 1

displays the macro SAM for the year 2001.

7 Mainly « Les Enquêtes Agricoles de base », « Annuaire des statistiques agricoles » and « Enquête sur les structures des exploitations agricoles ».

23

TABLE 1. THE 2001 MACRO SAM FOR TUNISIA (MILLION OF TD)

Activities Commodities Factors Institutions Fiscal Instruments SAV TOT AGR AGRF WAT MIN MANUF NMAN SERV AGRC AGRFC WATC MINC MANUFC NMANC SERC LAB CAP HS ENTR GOV ROW DTAX ITAX TIMP

AGR 4493.3 4493.3 AGRF 5843.4 5843.4 WAT 170.5 170.5 MIN 393.3 393.3 MAN 16500.9 16500.9 NMAN 7458.9 7458.9 SERV 18019.6 18019.6

AGRC 206.1 2417.5 3.2 126.8 2.1 209.4 2033.9 185.0 209.4 5393.5 AGRFC 477.3 922.3 65.8 1.3 664.9 3859.9 534.1 -0.4 6525.1 WATC 17.3 7.0 1.4 1.9 17.3 9.3 32.8 83.5 170.5 MINC 8.5 0.5 362.2 0.0 8.1 3.4 79.8 6.4 469.0 MANC 103.3 573.6 13.1 32.2 9005.6 2318.6 945.4 5588.8 7622.9 3198.6 29402.1 NMANC 91.5 138.1 14.6 44.0 749.3 939.6 762.3 765.1 892.9 4405.8 8803.1 SERVC 53.5 179.7 22.6 64.8 948.3 806.5 2689.8 4947.2 4745.3 4578.0 83.9 19119.4

LAB 508.7 525.4 63.3 110.7 2299.1 729.3 5958.3 69.6 10264.3

CAP 3033.9 460.3 37.3 135.0 2500.3 1920.5 6206.2 14293.5

HS 10201.1 8929.9 1402.3 1757.6 1464.1 23755.0 ENTR 5363.6 850.0 6.8 244.5 6464.9 GOV 2087.1 855.9 94.0 1893.4 2332.4 1686.1 8948.9 ROW 772.3 497.2 70.2 11603.8 1273.4 1099.8 63.2 101.0 657.5 902.9 17041.2

DTAX 1160.2 672.8 33.7 26.6 1893.4 ITAX 1.8 611.0 18.3 0.9 426.2 731.7 542.5 2332.4 TIMP 128.0 184.5 5.5 1297.4 70.7 1686.1 SAV 2275.0 2876.4 1502.6 1249.8 7903.7 TOT 4493.3 5843.4 170.5 393.3 16500.9 7458.9 18019.6 5393.5 6525.1 170.5 469.0 29402.1 8803.1 19119.4 10264.3 14293.5 23755.0 6464.9 8948.9 17041.2 1893.4 2332.4 1686.1 7903.7

24

VI. Main Estimation Results VI.1 Discrete choice model

The model that is taken to the data is given by (8). This equation represents a static panel data

model where we need to control for unobserved individual heterogeneity and simultaneity. To

alleviate endogeneity biases we use the Instrumental Variable estimator of Hausman and

Taylor (1980). Table 2 shows results from the OLS estimation and the IV specification.

TABLE 2. DEMAND EQUATION VARIABLES OLS IV Price Sj/g Constant

-0.135** (-3.55) 0.158** (12.87) -1.237** (-6.83)

-0.254** (-4.26) 0.326** (13.28) -1.15** (-6.7)

Number of Observations R²

6944 0.93

6944 0.95

Numbers in (.) are t-statistics. The significance at the 10% and 1% levels is indicated by * and ** respectively.

The coefficient of prices is significantly negative in both models, as wished. The IV results

show a larger prices coefficient (in absolute value) than the OLS counterpart, being consistent

with the expected correlation between prices and unobserved product quality that biases the

OLS estimates towards zero. The coefficient for conditional market shares is significantly

positive, and also larger for IV estimates. This shows a relatively important variability of the

consumer indirect utilities across the different groups of product.

Table 3 summarizes the mean values and standard deviations of the quality measures for the

main trading partners and competitors of Tunisia and for the main exchanged agricultural

commodities.

TABLE 3. QUALITY ESTIMATES Fruits Citrus Shell Fruits Vegetables Cereals Pulses Algeria Spain France Greece Italy Portugal

3.93 (0.18) 3.92 (0.2) 3.75 (0.17) 3.73 (0.26) 3.83 (0.15) 3.22

3.56 (0.18) 4.19 (0.13) 3.71 (0.16) 4.06 (0.11) 4.32 (0.14) 3.7

1.16 (0.21) 3.63 (0.49) 3.97 (0.08) 2.79 (0.85) 3.55 (0.34) 3.26

4.25 (0.31) 4.1 (0.07) 4.1 (0.08) 4.21 (0.24) 3.86 (0.13) 3.95

2.55 (0.57) 4.4 (0.37) 4.48 (0.31) 4.23 (0.49) 4.28 (0.41) 3.15

2.79 (0.8) 4.13 (0.66) 4.83 (0.46) 3.75 (0.56) 4.61 (0.6) 3.6

25

Israel Jordan Lebanon Morocco Syria Tunisia Turkey Egypt

(0.15) 3.92 (0.26) 4.12 (0.45) 3.94 (0.2) 4.09 (0.4) 5.16 (0.7) 5.12 (0.76) 4.57 (0.56) 3.35 (0.6)

(0.24) 3.82 (0.14) 4.42 (0.25) 4.47 (0.15) 4.05 (0.26) 5.07 (0.67) 4.95 (0.79) 4.86 (0.72) 2.93 (0.79)

(0.95) 3.15 (0.91) 4.77 (0.95) 5.1 (1.2) 3.25 (0.91) 5.2 (1.41) 4.42 (1.76) 5.63 (0.6) 3.11 (1.64)

(0.28) 4.1 (0.27) 4.11 (0.15) 4.12 (0.33) 3.93 (0.24) 4.95 (0.8) 5.12 (0.62) 5.18 (0.8) 3.88 (0.7)

(0.62) 3.52 (0.47) 3.65 (0.47) 3.82 (0.48) 3.94 (0.7) 5.16 (0.95) 3.95 (0.8) 4.57 (0.7) 5.06 (0.9)

(0.63) 3.1 (0.76) 3.4 (0.52) 0.31 (0.63) 3.41 (0.46) 4.1 (0.53) 2.11 (0.52) 4.32 (0.89) 2.15 (0.75)

a: Mean, b: Std. Dev. Averages are for the period : 1990 to 2005 VI.2. The latent class model

Table 4 presents the results of estimating the input elasticities of the production frontier in

equation (9).

TABLE 4: LATENT CLASS MODEL PARAMETER ESTIMATES: TOTAL POOL

CLASS 1 CLASS 2 CLASS 3 CLASS4 Production Frontier

Land Water Labor Fertilizers Machines Time Intercept

0.309*** 0.275*** 0.236*** 0.107* 0.097* 0.017*** 0.55**

0.261*** 0.289*** 0.26*** 0.092* 0.16* 0.06** 0.76**

0.444*** 0.276*** 0.141* 0.127* 0.136** 0.009** 0.022

0.216*** 0.333*** 0.144** 0.111* 0.327*** 0.008* 0.12

Efficiency term Land GINI Land fragmentation Land quality International trade Human capital R&D Gov. effectiveness γ= σu²/σ²

0.212*** 0.038** -0.04** -0.157*** -0.095*** -0.004* -0.026 0.72***

0.169*** 0.002* -0.04* -0.135*** -0.098** -0.002* -0.0034* 0.829***

0.175*** 0.058** -0.05*** -0.268*** -0.156** -0.002** -0.01** 0.784***

0.123*** 0.02* -0.011* -0.165*** -0.149** 0.001* 0.003*** 0.891***

Probabilities Total fertilizers Total machinery Total agricul. land Average farm size Rain Intercept

-0.073 0.079* 0.0367*** -0.026** -0.006* -1.36

0.144** -0.03 0.045** 0.35* 0.01** -1.359*

-0.99** 0.472*** 0.408*** 0.093** 0.262** -3.29**

26

Log-likelihood Number of Obs.

-274.33 1344

Notes: the variables in the production frontier and efficiency function are in natural logarithm. The significance at the 10%, 5% and 1% levels is indicated by *, ** and *** respectively. A negative sign in the inefficiency model means that the associated variable has a positive effect on technical efficiency. The average efficiency scores and TFP changes, estimated using equations (14) and (15)

respectively, are reported in Table 5. The results show productivity increases in the

Mediterranean agricultural sector, on average, with SMC registering relatively better average

rates of productivity gain than EU countries. Significant differences in technical efficiency

performance are apparent among commodity groups and countries. On average over the

period under consideration, EU countries exhibited better efficiency levels than SMC.

TABLE 5. LATENT CLASS MODEL AGRICULTURAL EFFICIENCY SCORES AND TFP GROWTH FRUITS CITRUS SHELL VEGETABLES CEREALS PULSES

TEa TFPGb TE TFPG TE TFPG TE TFPG TE TFPG TE TFPG ALGERIA EGYPT FRANCE GREECE ISRAEL ITALY JORDAN LEBANON MOROCCO PORTUGAL SPAIN SYRIA TUNISIA TURKEY

0.471 0.545 0.733 0.508 0.521 0.633 0.368 0.702 0.379 0.497 0.615 0.362 0.479 0.587

4.77 9.28 6.45 3.78 3.85 8.77 4.38 9.1

-0.71 0.76 6.19 2.96 1.82 9.42

0.375 0.718 0.691 0.787 0.741 0.777 0.565 0.675 0.712 0.716 0.882 0.739 0.559 0.942

4.82 4.89 -1.89 1.42 2.79 5.89 3.05 2.96 4.7

0.28 5.02 5.75 3.25 9.63

0.613 0.595 0.858 0.524 0.743 0.696 0.582 0.754 0.707 0.809 0.646 0.825 0.708 0.941

-1.52 1.71 1.13 -2.01 2.59 4.28 3.63 7.91 2.47 5.85 -1.96 5.85 2.55 8.77

0.403 0.352 0.592 0.413 0.607 0.511 0.689 0.865 0.428 0.743 0.539 0.616 0.566 0.642

-0.43 5.82 7.48 -0.48 3.83 6.45 3.08 9.88 6.29 -0.22 6.26 5.16 3.37 7.91

0.452 0.529 0.902 0.636 0.397 0.656 0.283 0.508 0.481 0.509 0.641 0.619 0.527 0.771

3.98 5.07 6.57 4.16 -1.24 6.26 -1.65 7.88 1.49 1.6

9.48 3.49 1.23 5.92

0.521 0.624 0.953 0.611 0.629 0.709 0.743 0.847 0.561 0.493 0.537 0.669 0.557 0.737

-0.79 4.89 6.4

0.97 4.02 1.11 3.47 -1.59 4.56 -0.92 6.1

1.27 2.41 8.27

a: Technical efficiency score, b: TFP growth.

Table 6 reports the estimation results of equation (16) considering three measures of

international trade, namely agricultural equipment imports share (column 1), the ratio of

agricultural exports plus imports to AGVA (column 2), and agricultural trade barriers

(column 3).

TABLE 6: IMPACT OF INTERNATIONAL TRADE ON AGRICULTURAL TFP GROWTH MACHINERY

IMPORTS TRADE

VOLUMES TRADE

BARRIERS Human capital International Trade GAP H*GAP IT*GAP Land GINI Land fragmentation Land quality R&D Cont. of Corruption

0.065** 0.912*** -0.146*** -0.085**

-0.174*** -0.026** -0.046 0.058* 0.003* 0.007*

0.014** 0.232*** -0.175*** -0.247** -0.121** -0.02* -0.032* 0.052* 0.003* 0.033*

0.09** -0.424*** -0.186*** -0.157*** 0.163*** -0.058***

-0.032* 0.031* 0.011* 0.004*

27

Gov. effectiveness Political stability

0.03*** 0.004

0.014** 0.004

0.016* 0.002

N. of observations R² adjusted

1260 0.75

1260 0.73

1260 0.71

Notes: Dependant variable is GTFP. H*Gap: the product of human capital (H) and technology gap (GAP),

IT*GAP: the product of international trade (IT) and technology gap. H, IT, GAP and the interaction terms are

instrumented by their second lags. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels

respectively.

Regardless of the international trade measure, the results lend strong support to the positive

effect of trade openness on agricultural productivity growth. Across the three regressions,

TFP growth rate increases with higher trade shares and decreases with more trade barriers.

The trade interaction is negative and statistically significant at the 1% level in all the

regressions, indicating that the further a country lies behind the frontier and the greater the

potential for trade to increase the rates of agricultural TFP growth through the speed of

technology transfer. As foreign technology diffuses mainly through capital goods, the trade

effects seem to be better captured by agricultural equipment imports share. Notice also, that

the technology gap variable enters negatively and is statistically significant, supporting the

existence of a catch up effect. The estimated coefficients on the human capital level and

interaction terms indicate a small direct impact of educational attainment on the rates of TFP

growth, but a relatively important effect on the rate of technology transfer. This finding is

consistent with the notion that countries with sufficient educational attainment benefit

positively from advanced technology brought along by international activities.

These estimates provide interesting insights into the agricultural productivity dynamics. The

results highlight the role of international trade in promoting technology transfer and point to

the importance of education in facilitating the assimilation of foreign improvement of

technology. The findings suggest that international trading opportunities would have larger

benefits in countries with favourable internal factors relating to more equitable distributions

of land, better land quality, significant R&D and positive institutional conditions.

VII. Simulation of trade policy reform

In this section we evaluate three sets of scenarios:

Scenario 1: Cutting tariffs on manufactured products imported from the European

Union.

Scenario 2: In addition to scenario 1 this simulation assumes 25% decrease of tariff

barriers on agricultural and agro-food imports from the European Union.

28

Scenario 3: This scenario extends Scenario 1 to agricultural and agro-food products

and the non-EU countries.

The simulation analysis focuses only on selected key variables, the choice of which relies on

the mechanisms through which trade policy affects inequality and poverty. The results are

reported using the percentage deviation from the model’s base-line:

TABLE 7. PRODUCTIVITY AND QUALITY Commodities Total Factor Productivity Labor Productivity Quality of Commodity

Scenario 1 Scenario 2 Scenario 3 Scenario 1 Scenario 2 Scenario 3 Scenario 1 Scenario 2 Scenario 3

TWHEAT 0.07 0.27 1.67 0.13 0.18 0.29 -2.33 98.36 107.67 HWHEAT 0.15 0.33 1.52 0.28 0.55 2.16 -3.01 98.18 112.78 BARLEY 0.13 0.22 0.50 0.21 0.29 0.52 -3.63 -4.17 -2.57 OCER 0.26 0.34 0.74 0.30 0.39 0.74 0.86 1.37 4.60 LEGUM 0.20 0.26 5.76 0.46 0.59 9.25 -0.24 -0.23 74.84 OLIV 0.17 0.21 0.30 0.30 0.38 0.54 0.39 0.51 1.45 CITR 0.28 0.37 0.57 0.79 1.02 1.58 0.94 1.26 2.67 DAT 0.27 0.35 0.54 0.88 1.14 1.73 0.91 1.19 2.73 OFRUITS 0.18 0.55 6.89 0.16 -0.67 -18.37 -0.27 12.34 266.17 VEG 0.21 0.31 0.42 0.46 0.68 0.92 -0.08 0.48 0.99 LVST 0.22 0.51 1.61 0.88 2.00 6.31 -0.02 2.93 16.61 INDCUL 0.17 0.22 0.49 0.25 0.11 -3.62 1.04 1.77 7.85 OCROPS 0.16 0.22 0.35 0.20 0.23 0.24 -1.78 0.04 10.09 FISH 0.25 0.35 0.74 2.37 3.32 7.01 0.39 1.05 5.80 MEAT 0.24 0.38 0.71 0.82 1.26 2.37 0.31 1.50 6.17 DAIRY -0.04 0.49 6.35 -0.05 0.67 8.63 -1.70 4.88 79.05 FLOUR 0.22 0.68 3.66 0.34 1.04 5.64 0.38 6.35 48.88 OOIL 0.16 0.20 0.31 0.50 0.63 0.96 0.11 0.14 0.56 OGR 0.23 0.34 0.97 0.73 1.06 3.03 0.83 1.62 7.69 CANNED 0.23 0.50 2.69 0.52 1.12 6.07 0.88 3.76 30.71 SUGAR 0.09 0.31 1.34 0.14 0.51 2.19 -0.41 2.06 15.43 BEVER 0.16 0.66 3.13 0.35 1.38 6.57 -0.45 5.66 38.03 OAGRF 0.16 0.26 3.68 0.22 0.36 5.15 -0.10 0.79 43.65 MCV 1.10 1.68 1.71 2.11 3.20 3.26 7.82 11.97 9.80 IME 0.23 0.30 0.34 0.42 0.53 0.61 -0.41 -0.58 -2.51 CHEM 0.49 0.66 0.76 1.08 1.45 1.68 3.62 4.66 4.57 TEXT 0.28 0.36 0.48 0.49 0.62 0.83 1.84 2.34 2.38 OMAN 0.97 1.28 1.39 1.32 1.74 1.88 7.98 9.81 9.51 MINING 0.37 0.91 1.00 0.87 2.14 2.38 1.30 4.54 4.39 WATER 0.16 0.21 0.29 0.22 0.29 0.41 0.00 0.00 0.00 NMAN 0.16 0.23 0.31 0.42 0.61 0.81 -3.93 -4.57 -6.97 SERV 0.25 0.32 0.45 0.45 0.58 0.82 0.43 0.60 0.86

TABLE 8. SUPPLY AND DEMAND

TOTAL PRODUCTION DOMESTIC DEMAND COMPOSITE PRICE Scenario 1 Scenario 2 Scenario 3 Scenario 1 Scenario 2 Scenario 3 Scenario 1 Scenario 2 Scenario 3

TWHEAT 0.06 -0.98 -9.19 0.06 -0.98 -9.19 -0.26 -2.07 -14.88 HWHEAT -2.03 -2.18 0.20 -2.03 -2.18 0.20 -0.71 -1.74 -8.76 BARLEY -2.62 -3.42 -3.66 -2.88 -3.81 -4.35 -0.24 -0.85 -3.93 OCER 1.31 1.91 1.85 1.36 2.00 1.68 -0.03 -0.06 -4.13 LEGUM 0.87 1.28 -13.34 0.87 1.28 -13.96 -0.47 -0.63 -36.07 OLIV 1.68 2.23 3.25 1.68 2.23 3.25 0.27 0.36 -0.51 CITR 0.72 0.99 1.65 0.66 0.91 1.45 -1.03 -1.39 -3.21 DAT 0.84 1.13 2.28 0.66 0.90 1.60 -0.99 -1.31 -3.23 OFRUITS 0.69 -0.24 -24.52 0.68 -0.27 -25.04 -0.89 0.23 27.76 VEG 0.66 0.96 2.01 0.66 0.96 2.00 -0.46 -0.72 -2.64 LVST 0.47 0.90 3.22 0.47 0.89 3.14 -0.58 -1.39 -7.69

29

INDCUL 2.69 0.82 -46.61 2.80 0.71 -51.34 -0.07 -2.06 -36.52 OCROPS 0.18 -0.08 -0.88 0.14 -0.16 -1.09 -1.66 -2.64 -4.84 FISH 0.38 0.70 2.51 0.39 0.71 2.38 0.23 -0.03 -3.81 MEAT 0.61 1.23 4.16 0.61 1.23 4.16 -0.47 -1.27 -6.79 DAIRY 2.55 3.79 8.65 2.55 3.75 8.22 -2.46 -5.95 -35.14 FLOUR 1.08 2.71 12.47 1.08 2.60 11.52 -0.89 -4.00 -25.57 OOIL 1.69 2.25 3.23 5.65 7.60 9.69 2.60 3.56 3.92 OGR 1.49 1.98 1.89 1.53 1.99 1.29 -0.11 -1.27 -15.41 CANNED 1.56 2.24 4.23 1.78 1.43 -8.13 -0.67 -3.90 -35.16 SUGAR 2.22 3.02 6.14 2.23 3.00 5.97 -0.99 -3.56 -18.21 BEVER 0.88 2.15 6.83 0.84 1.85 5.24 -1.34 -6.31 -29.75 OAGRF 1.46 2.22 5.31 1.40 2.11 4.10 -2.11 -3.38 -20.37 MCV -3.64 -6.08 -7.27 -5.32 -8.15 -9.43 -14.83 -20.32 -20.39 IME -0.20 -0.26 -1.40 -3.61 -4.37 -5.97 -8.07 -10.33 -10.20 CHEM 0.86 0.81 1.14 -2.45 -3.60 -3.27 -8.67 -12.14 -12.23 TEXT 2.03 2.63 2.89 1.48 1.97 2.22 -2.46 -3.16 -3.26 OMAN -1.43 -2.51 -2.45 -2.77 -4.13 -4.08 -11.73 -15.38 -15.46 MINING -0.13 -3.35 -2.99 -0.44 -4.09 -3.77 -1.57 -2.20 -2.55 WATER 0.81 1.07 1.02 0.81 1.07 1.02 -1.78 -2.40 -2.71 NMAN -3.65 -4.43 -6.67 -4.63 -5.65 -8.18 -6.47 -8.69 -9.83 SERV 0.72 1.00 1.40 0.62 0.87 1.15 -0.88 -1.23 -1.96

TABLE 9. HOUSEHOLD CONSUMPTION

HOUSEHOLD CONSUMPTION RURAL URBAN Scenario 1 Scenario 2 Scenario 3 Scenario 1 Scenario 2 Scenario 3 TWHEAT 0.67 1.80 9.23 0.23 1.08 6.13 HWHEAT 0.87 1.65 6.47 0.44 0.93 3.40 BARLEY 0.00 0.00 0.00 0.00 0.00 0.00 OCER 0.56 0.89 4.37 0.13 0.18 1.33 LEGUM 0.76 1.15 18.78 0.33 0.43 15.58 OLIV 0.00 0.00 0.00 0.00 0.00 0.00 CITR 1.03 1.49 3.96 0.58 0.77 0.92 DAT 1.01 1.45 3.97 0.57 0.74 0.93 OFRUITS 0.96 0.76 -10.01 0.52 0.05 -12.90 VEG 0.76 1.19 3.70 0.32 0.48 0.67 LVST 0.81 1.49 5.98 0.38 0.77 2.92 INDCUL 0.00 0.00 0.00 0.00 0.00 0.00 OCROPS 1.32 2.05 4.70 0.87 1.33 1.65 FISH 0.70 1.38 6.65 0.02 0.26 1.87 MEAT 1.20 2.26 8.76 0.52 1.13 3.96 DAIRY 2.67 5.57 28.85 1.95 4.41 23.84 FLOUR 1.50 4.19 22.07 0.81 3.04 17.13 OOIL -0.95 -1.17 1.17 -1.60 -2.26 -3.55 OGR 0.94 2.26 14.87 0.26 1.13 10.00 CANNED 1.35 4.12 28.87 0.66 2.97 23.86 SUGAR 1.58 3.88 16.85 0.88 2.73 11.97 BEVER 1.83 5.82 25.03 1.13 4.66 20.06 OAGRF 2.41 3.75 18.38 1.70 2.61 13.48 MCV 13.36 15.75 18.40 12.43 14.49 13.50 IME 7.16 8.68 11.18 6.35 7.48 6.35 CHEM 7.67 9.96 12.62 6.86 8.75 7.78 TEXT 2.67 3.59 6.26 1.96 2.45 1.48 OMAN 10.39 12.25 14.91 9.52 11.02 10.04 MINING 2.00 2.92 5.76 1.30 1.78 0.99 WATERNA 2.16 3.06 5.87 1.45 1.92 1.10 WATERA 0.56 1.35 12.05 0.00 0.00 0.00 NMAN 5.83 7.51 10.92 5.05 6.33 6.09 SERV 1.43 2.13 5.10 0.77 1.05 0.55

30

TABLE 10. EXTERNAL TRADE EXPORT SUPPLY EXPORT DEMAND IMPORT DEMAND EU ROW EU ROW EU ROW Scen. 1 Scen. 2 Scen. 3 Scen. 1 Scen. 2 Scen. 3 Scen. 1 Scen. 2 Scen. 3 Scen. 1 Scen. 2 Scen. 3 Scen. 1 Scen. 2 Scen. 3 Scen. 1 Scen. 2 Scen. 3 TWHEAT 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -3.05 1.10 8.68 -3.05 -6.03 12.04 HWHEAT 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -3.94 0.65 17.38 -3.94 -5.67 16.90 BARLEY 0.00 0.00 0.00 -0.94 -0.97 0.70 0.00 0.00 0.00 -0.94 -0.97 0.70 -4.92 -4.99 -3.56 -4.92 -6.97 -3.59 OCER 0.00 0.00 0.00 0.76 1.11 3.47 0.00 0.00 0.00 0.76 1.11 3.47 1.16 4.16 6.34 1.16 1.76 6.19 LEGUM 0.00 0.00 0.00 0.78 1.11 24.69 0.00 0.00 0.00 0.78 1.11 24.69 0.00 0.00 0.00 -0.37 -0.38 102.34 OLIV 0.51 0.68 1.91 0.51 0.68 1.91 0.51 0.68 1.91 0.51 0.68 1.91 0.00 0.00 0.00 0.00 0.00 0.00 CITR 1.24 1.66 3.53 1.24 1.66 3.53 1.24 1.66 3.53 1.24 1.66 3.53 -1.41 -1.87 -4.97 -1.41 -1.87 -4.97 DAT 1.20 1.58 3.62 1.20 1.58 3.62 1.20 1.58 3.62 1.20 1.58 3.62 -1.33 -1.73 -4.86 -1.33 -1.73 -4.86 OFRUITS 0.94 1.58 8.33 0.94 1.58 8.33 0.94 1.58 8.33 0.94 1.58 8.33 -1.14 25.26 560.51 0.00 0.00 0.00 VEG 0.62 1.03 3.02 0.62 1.03 3.02 0.62 1.03 3.02 0.62 1.03 3.02 -0.29 0.53 0.84 0.00 0.00 0.00 LVST 0.61 1.64 8.43 0.61 1.64 8.43 0.61 1.64 8.43 0.61 1.64 8.43 -0.71 7.64 36.37 -0.71 -1.64 35.61 INDCUL 0.00 0.00 0.00 1.82 1.62 -12.00 0.00 0.00 0.00 1.82 1.62 -12.00 1.36 18.68 11.68 1.36 -1.75 10.41 OCROPS 1.22 2.02 4.34 1.22 2.02 4.34 1.22 2.02 4.34 1.22 2.02 4.34 -3.54 -0.08 16.33 -3.54 -5.42 15.94 FISH 0.05 0.52 5.17 0.05 0.52 5.17 0.05 0.52 5.17 0.05 0.52 5.17 0.88 4.17 9.78 0.88 0.73 9.55 MEAT 0.61 1.55 7.21 0.61 1.55 7.21 0.61 1.55 7.21 0.61 1.55 7.21 -0.34 3.53 11.44 0.00 0.00 0.00 DAIRY 2.83 6.20 33.39 2.83 6.20 33.39 2.83 6.20 33.39 2.83 6.20 33.39 -2.98 13.78 115.53 -2.98 -7.19 112.89 FLOUR 1.12 4.44 26.92 1.12 4.44 26.92 1.12 4.44 26.92 1.12 4.44 26.92 -0.76 16.34 140.63 0.00 0.00 0.00 OOIL 0.15 0.19 0.74 0.15 0.19 0.74 0.15 0.19 0.74 0.15 0.19 0.74 0.00 0.00 0.00 0.00 0.00 0.00 OGR 0.81 1.82 11.11 0.81 1.82 11.11 0.81 1.82 11.11 0.81 1.82 11.11 1.12 8.17 10.44 1.12 -0.42 9.88 CANNED 1.34 3.05 16.49 1.34 3.05 16.49 1.34 3.05 16.49 1.34 3.05 16.49 0.19 20.29 174.68 0.19 -3.28 171.09 SUGAR 1.83 4.00 15.67 1.83 4.00 15.67 1.83 4.00 15.67 1.83 4.00 15.67 -0.63 5.97 20.73 -0.63 -4.05 19.99 BEVER 1.36 5.72 25.94 1.36 5.72 25.94 1.36 5.72 25.94 1.36 5.72 25.94 -2.08 8.75 68.37 0.00 0.00 0.00 OAGRF 2.20 3.44 19.06 2.20 3.44 19.06 2.20 3.44 19.06 2.20 3.44 19.06 -3.09 13.47 108.24 -3.09 -4.75 106.03 MCV 10.05 11.53 11.07 10.05 11.53 11.07 10.05 11.53 11.07 10.05 11.53 11.07 30.42 18.19 14.10 -29.30 17.20 13.15 IME 4.41 5.41 4.92 4.41 5.41 4.92 4.41 5.41 4.92 4.41 5.41 4.92 3.98 -2.30 -5.38 -17.45 -2.63 -5.71 CHEM 4.71 6.03 6.36 4.71 6.03 6.36 4.71 6.03 6.36 4.71 6.03 6.36 13.20 6.31 5.93 -15.11 5.86 5.48 TEXT 2.42 3.09 3.36 2.42 3.09 3.36 2.42 3.09 3.36 2.42 3.09 3.36 3.83 3.09 2.89 -2.96 2.99 2.78 OMAN 7.28 8.31 8.44 7.28 8.31 8.44 7.28 8.31 8.44 7.28 8.31 8.44 20.13 14.14 13.61 -20.09 13.47 12.94 MINING 0.86 -0.99 -0.52 0.86 -0.99 -0.52 0.86 -0.99 -0.52 0.86 -0.99 -0.52 13.00 12.05 11.28 -2.79 11.79 11.02 WATERNA 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 WATERA 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 NMAN 3.22 4.29 4.10 3.22 4.29 4.10 3.22 4.29 4.10 3.22 4.29 4.10 -8.08 -15.37 -21.21 -17.48 -15.51 -21.34 SERV 0.99 1.37 2.08 0.99 1.37 2.08 0.99 1.37 2.08 0.99 1.37 2.08 -1.28 -1.79 -3.09 -1.28 -1.79 -3.09

31

Bibliography

[1] Anderson K. and Martin W. (2005) “Agricultural Trade Reform and the Doha

Development Agenda” World Bank, WPS 3607.

[2] Annabi N., Cissé F., Cockburn J. and Decaluwé B. (2005) “Trade Liberalisation, Growth

and Poverty in Senegal: a Dynamic Microsimulation CGE Model Analysis” Cahier de

recherche/Working Paper 05-12, CIRPÉE.

[3] Battese, G.E. and D.S.P. Rao (2002) “Technology Gap, Efficiency and a Stochastic

Metafrontier Function”, International Journal of Business and Economics, 1 (2).

[4] Battese, G. D.S.P. Rao, and C.J. O’Donnell (2004) “A metafrontier production function

for estimation of technical efficiencies and technology gaps for firms operating under

different technologies”, Journal of Productivity Analysis 21, 91-103.

[5] Bchir M.H., Decreux Y., Guérin J.-L. and S. Jean. 2002. ‘Mirage, a Computable General

Equilibrium Model for Trade Policy Analysis’, CEPII Working Paper, 2002-17, CEPII,

Paris.

[6] Benhabib J. and Spiegel M. (2003) “Human Capital and Technology Diffusion”,

Handbook of Economic Growth.

[7] Berry S.T.(1994) ''Estimating Discrete-Choice Models of Product Differentiation'', RAND

Journal of Economics, 25, p. 242-262.

[8] Bernard, A. B., Stephen J. R. and Schott P.K. (2006) “Comparative Advantage and

Heterogeneous Firms”, Review of Economic Studies, forthcoming.

[9] Bibi S. and Chatti R. (2006) “Trade Liberalization and the Dynamics of Poverty in

Tunisia: A Layered CGE Microsimulation Analysis” MPIA, Working paper 2006-07.

[10] Bigsten A., Collier P., Dercon S., Fafchamps M., Gauthier B., Gunning J.W., Oduro A.,

Oostendorp R., Pattillo C., Soderbom M., Teal F. and Zeufack A. (2004) “Do African

Manufacturing Firms Learn From Exporting”, The Journal of Development Studies, 40,

p.115-141.

[11] Bouët A., Decreux Y., Fontagné L., Jean S. and D. Laborde (2004) A consistent, ad-

valorem equivalent measure of applied protection across the world: the MacMap-HS6

database” CEPII Working Paper.

[12] Boughzala, M., H. Bchir, S. Bibi, R. Chatti and T. Rajhi (2005) “Trade, Employment and

Wages in Tunisia: An Integrated and Dynamic CGE Model” Research n°FEM21-29,

FEMISE RESEARCH PROGRAM, October 2005.

32

[13] Cameron, G., Proudman, J., and Redding, S. (2005). Technological Convergence, R&D,

Trade and Productivity Growth. European Economic Review 49: 775 – 807.

[14] Caudill S. (2003) “Estimating a Mixture of Stochastic Frontier Regression Models via the

EM algorithm: A multiproduct Cost Function Application” Empirical Economics 28,

p.581-598.

[15] Chemengui M.A. and Dessus S. (1999) “ La Libéralisation de l’Agriculture Tunisienne et

l’Union Européenne : Une vue prospective ». OCDE, Document technique, 1999.

[16] Chemengui M.A. and Thabet C. (2006) “Agricultural Trade Liberalization and Poverty

in Tunisia: Micro-Simulation in a General Equilibrium Framework” MPIA Network

Session Paper, 5th PEP Research Network General Meeting, June 2006, Addis Ababa

Ethiopia.

[17] Coxhead, Ian and Peter Warr. 1995. “Does technical progress in agriculture alleviate

poverty? A Philippine case study.” Australian J. Agr. Econ., 39(1), April, pp.25-54.

[18] Cline, William. 2004. Trade Policy and Global Poverty. Washington, D.C.: Institute for

International Economics.

[18] Decaluwé B., Patry, A. Savard, L. and E. Thorbecke. (1999), Poverty Analysis within a

General Equilibrium Framework. Cahier de recherche 99-09, CREFA, Université Laval,

Quebec.

[19] De Janvry and E. Sadoulet (2001) “World Poverty and the Role of Agricultural

Technology: Direct and Indirect Effects” Journal of Development Studies.

[20] Diao, X, Rattso J, and Stokke, H.E. (2005) “International spillovers, productivity growth

and openness in Thailand: an intertemporal general equilibrium analysis” Journal of

Development Economics 76 (2005) 429– 450

[21] ESCWA (2001) “Agricultural Trade and the New Trade Agenda: Options and Strategies

to Capture the Benefits for the Middle East, Case Study from TUNISIA”, ECONOMIC

AND SOCIAL COMMISSION FOR WESTERN ASIA.

[22] Feenstra R. C. and Kee H.L. (2004) “Export Variety and Country Productivity” NBER

Working Paper 10830.

[23] Greene W. (2001) “New Developments in the Estimation of Stochastic Frontier Models

with Panel Data”, Efficiency Series Paper 6/2001, Dpto. de Economía, Univ. de Oviedo.

[24] Greene, W. (2005) “Reconsidering heterogeneity in panel data estimators of the stochastic

frontier model” Journal of Econometrics 126, p. 269–303.

33

[25] Griffith, R., Redding, S., and Van Reenen, J. (2004). Mapping the Two Faces of R&D:

Productivity Growth in a Panel of OECD Industries. The Review of Economics and

Statistics 86(4): 883–895.

[26] Hallak J.C. (2003) “The Effect of Cross-Country Differences in Product Quality on the

Direction of International Trade” RSIE Working paper 493.

[27] Hallak J.C. (2005) “Product Quality and the Direction of Trade” Forthcoming in Journal

of International Economics.

[28] Hallak J.C. and Scott P.K. (2005) “Estimating Cross-Country Differences in Product

Quality” Yale University, mimeo.

[29] Hummels D. and Klenow P. (2005) “The Variety and Quality of a Nation’s Export”

American Economic Review 95, 704-723

[30] Khandelwal A. (2005) “Product Quality and Competition in International Trade” Yale

University, mimeo.

[31] Koetter, M., Bos J., Economidou C, and Kolari J. (2007) “Do Technology and Efficiency

Differences determine Productivity?” Discussion Paper Series No. 07-14, Utrecht School

of Economics, Tjalling C. Koopmans Research Institute.

[32] Kraay A., Soloaga I. and Tybout J. (2002) “Product Quality, Productive Efficiency, and

International Technology Diffusion: Evidence from Plant Level Panel Data” Policy

Research Working Paper, The World Bank, 2002.

[33] Kumbhakar S.C. and Tsionas E. (2003) ”Recent Developments in Stochastic Frontier

Modelling”, Efficiency Series Paper 6/2003, Departemento de Economía. Universidad de

Oviedo.

[34] Kumbhakar S.C. (2004) “Productivity and Efficiency Measurement Using Parametric

Econometric Methods” XIII International Tor Vergata Conference on Banking and

Finance Rome, Italy (Dec 1-3, 2004).

[35] Melitz M. J. (2001) “Estimating Firm-Level Productivity in Differentiated Product

Industries”, Harvard mimeo.

[36] Melitz M. J. (2003) “The Impact of Trade on Intra-Industry Reallocations and Aggregate

Industry Productivity”, Econometrica, Vol. 71, November 2003, p. 1695-1725.

[37] Nelson R. and Phelps E. (1966) “Investment in Humans, Technology Diffusion and

Economic Growth”. Papers and Proceedings-American Economic Review 56, 69– 75.

[38] Nicita A. (2003) “The Effects of Mexican Trade Liberalization on Household Welfare.”

World Bank, Washington, D.C.

34

[39] Porto G. (2003a) “Trade Reforms, Market Access and Poverty in Argentina.” World

Bank, Washington, D.C.

[40] Porto G. (2003b) “Using Survey Data to Assess the Distributional Effects of Trade

Policy.” World Bank, Washington, D.C.

[39] Rattsø, J and Stokke H.E. (2005) “Ramsey model of barriers to growth and skill-biased

income distribution in South Africa” Norwegian University of Science and Technology.

[40] Ravallion M. (1990) “Rural Welfare Effects of Food Price Changes Under Induced Wage

Responses: Theory and Evidence for Bangladesh.” Oxford Economic Papers 42: 574-585.

[41] Sánchez Cantillo, M.C. (2004) “Rising Inequality and Falling Poverty in Costa Rica's

Agriculture during Trade Reform: A Macro-micro General Equilibrium Analysis” PHD

thesis, The Institute of Social Studies , the Hague, the Netherlands.

[42] Stokke H. (2004) “Technology Adoption and Multiple Growth Paths: an Intertemporal

General Equilibrium Analysis of the Catch-up Process in Thailand” Review of World

Economics/Weltwirtschaftliches Archiv 140 (1), 80– 109.

[43] Van Der Mensbrugghe D. (2005) “Prototype Model for a Single Country Real

Computable General Equilibrium Model” Development Prospects Group, THE WORLD

BANK.

[44] Verhoogen E. (2004) “Trade, Quality Upgrading, and Wage Inequality in the Mexican

Manufacturing Sector: Theory and Evidence from Exchange Rate Shock”, UC Berkley

Mimeo.

[45] Winters L.A. (2004), ‘Trade Liberalization and Economic Performance: An Overview’,

Economic Journal 114: F4-F21, February.

[46] Winters, A. (2005) “The European agricultural trade policies and poverty”, European

Review of Agricultural Economics Vol 32 (3) (2005) p. 319–346.

[47] Winters L. A., McCulloch N. and McKay A. (2004) “Trade Liberalization and Poverty:

The Evidence so far.” Journal of Economic Literature 42(1):72-115.

35

Appendix: THE GENERAL EQUILIBRIUM MODEL EQUATIONS I- Production

1. jjj XSioCI =

2. jjj XSVA ν=

3. Kaga

Daga

Laga

agaagaagaVAagaaga KDLATLDTAVA βββ=

4. agaagaLagaagaaga VAPVAWTLDT β=

5. agaagaDagaagaaga VAPVArdtLAT β=

6. jjkjjj VAPVArkKD β=

7. Lagr

Lagr

Lagr

Lagr

agrLagr

Lagr

agrLagr

1

agrfawLagragrfaw

agrswa2

biasLagragrswaagruwa

2bias

Lagragruwaagr

LDA

LDALDALDT

ρρρ

ρρρρ

γ

γγ

−

−−

−+−−−−

⎥⎦⎤+

⎢⎣

⎡+=

,'',''

,'','','',''

8. ( ) agruwabiasL

agruwa

swa

swa

uwaagrswa LDA

WW

LDLagr

agr

,''''

''

''

'',''

σ

γγ

⎥⎦

⎤⎢⎣

⎡=

9. ( ) agruwa2

biasLagr

uwa

faw

faw

uwaagrfaw LDA

WW

LD

Lagr

agr

,''''

''

''

'',''

σ

γγ

⎥⎥⎦

⎤

⎢⎢⎣

⎡=

10. ( )LDagaLD

aga

DagaLD

agaLDaga

DagaLD

aga

1

aga2

biasDaga

LDagaagaadal

2bias

Daga

LDagaaga WLANA1LANALAT

ρρρρρ γγ

−

−+−−−−

⎥⎥⎦

⎤

⎢⎢⎣

⎡−+= ,''

11. ( ) agaadalbiasD

agaLDaga

LDaga

aga

adalaga LANA

1rdw

rdagaWLAN

LDaga

Daga

,''''

σ

γγ

⎥⎥⎦

⎤

⎢⎢⎣

⎡ −=

12. ( )[ ] LWaga

DWaga

DWaga

1

agawateraDWagaagaaial

DWaga

DWagaaga DI1LANAWLAN ρρρ γγ

−−− −+= ,'',''

13. agawateraDWaga

DWaga

aial

wateralagaaial DI

1rdagaPCLAN

DWaga

,''''

'',''

σ

γγ

⎥⎥⎦

⎤

⎢⎢⎣

⎡

−=

14. ( )LDagpLD

agp

DagpLD

agpLDagp

DagpLD

agp

1

agp2

biasDagp

LDagpagppdal

2bias

Dagp

LDagpagp WLANA1LANALAT

ρρρρρ γγ

−

−+−−−−

⎥⎥⎦

⎤

⎢⎢⎣

⎡−+= ,''

36

15. ( ) agppdalbiasD

agpLDagp

LDagp

agp

agppdalagp LANA

1rdw

rdagpWLAN

LDagp

Dagp

,'',''

σ

γγ

⎥⎥⎦

⎤

⎢⎢⎣

⎡ −=

16. ( )[ ] LWagp

DWagp

DWagp

1

agpwateraDWagpagppial

DWagp

DWagpagp DI1LANAWLAN ρρρ γγ

−−− −+= ,'',''

17. agpwateraDWagp

DWagp

agppial

wateralagppial DI

1rdagpPCLAN

DWagp

,'',''

'',''

σ

γγ

⎥⎥⎦

⎤

⎢⎢⎣

⎡

−=

18. Knag

Lnag

nagnagVAnagnag KDLDTAVA ββ=

19. nagnagLnagnagnag VAPVAWTLDT β=

20. LagrL

nagnagL

nagLnag

nagLnag

1

nagswna2

biasLnagnagswnanaguwna

2bias

Lnagnaguwnanag LDALDALDT

ρρρρρ γγ

−

−+−−−−⎥⎦

⎤⎢⎣

⎡+= ,'','','',''

21. ( ) naguwnabiasL

naguwna

swna

swna

uwnanagswna LDA

WW

LDLnag

nag

,''''

''

''

'',''

σ

γγ

⎥⎦

⎤⎢⎣

⎡=

22. jjiji CIaijDI ,, =

II- Productivity

23. ( ) ( ) Dagr

Lagr D

agrLagr

VAagragr AAAA ββ

=

24. ( ) LnagrL

nagVAnagnag AAA β

=

25. ⎥⎦

⎤⎢⎣

⎡−

⎥⎥⎦

⎤

⎢⎢⎣

⎡⎥⎦⎤

⎢⎣⎡+⎥⎦

⎤⎢⎣⎡=

−Fj

j0j

jHj

H0j

0jj

AA

1XSP

TRADEGDP

GbGDP

GA

AAOP

2H1H ααα

αα

26. ⎥⎦

⎤⎢⎣

⎡−

⎥⎥⎦

⎤

⎢⎢⎣

⎡⎥⎦⎤

⎢⎣⎡=

−F

agr

agr0

agr

agrHDagr0D

agr

0Dagr

Dagr

AA

1XSP

TRADEGDP

GbA

AADOP

2DH αα

α

27. ⎟⎟⎟

⎠

⎞

⎜⎜⎜

⎝

⎛−

⎥⎥⎦

⎤

⎢⎢⎣

⎡= 1

XSTRADEXSTRADE

BIAS2

0j

0j

jjBjj /

/α

28. ⎟⎟⎟

⎠

⎞

⎜⎜⎜

⎝

⎛−

⎥⎥⎦

⎤

⎢⎢⎣

⎡= 1

XSTRADEXSTRADE

BIAS2

0agr

0agr

agragrBDagr

Dagr /

/α

37

III- Income and savings

29.

∑∑

∑ ∑∑∑ ∑

+++⎟⎟⎠

⎞⎜⎜⎝

⎛+

⎥⎥⎦

⎤

⎢⎢⎣

⎡⎟⎟⎠

⎞⎜⎜⎝

⎛+⎟⎟

⎠

⎞⎜⎜⎝

⎛+⎟⎟⎠

⎞⎜⎜⎝

⎛=

r

Hrh

Hhh

jjj

Kh

land agpagplandagpland

agaagalandland

Dlandh

l jjll

Llhh

eTRRTRGDIVKDrk

LANrdagpLANrdagaLDWYH

,

,,,,,,

λ

λλ

30. Ghhhh TRHDTHYHYDH −−=

31. hhh YDHpmsSH =

32. ∑−−−=r

Rhr

Fhhhh TRHTRHSHYDHCTH ,

33. ∑∑∑∑ +++⎟⎟⎠

⎞⎜⎜⎝

⎛⎟⎠

⎞⎜⎝

⎛−=

r

Fr

F

h

Fh

jjj

h

Kh eTRRTRGTRHKDrk1YF λ

34. ⎟⎟⎠

⎞⎜⎜⎝

⎛⎟⎠

⎞⎜⎝

⎛−= ∑∑

jjj

h

Kh

DIVhh KDrk1DIV λγ

35. ⎟⎟⎠

⎞⎜⎜⎝

⎛⎟⎠

⎞⎜⎝

⎛−= ∑∑

jjj

h

Kh

DIVRr

Rr KDrk1TRF λγ

36. G

r

Rr

hh TRFDTFTRFDIVYFSF −−−−= ∑∑

37. ( ) ∑∑∑ ++++++=r

rr

Gr

G

h

Ghh TIMTIeTRRTRFDTFTRHDTHYG

38. hHhh YHtdDTH =

39. hH

hGh YHtrTRH =

40. YFtdDTF F=

41. YFtrTRF FG =

42. ∑ ∑ ⎥⎦

⎤⎢⎣

⎡++=

j rrjrjrjjjjj IMtm1ePWMtxDPLtxTI ,,, )(

43. [ ]∑=j

rjrjrj IMePWMtmTIM ,,,

44. ∑∑ −−−−=r

Rr

F

h

Hh TRGTRGTRGGYGSG

IV- Demand

38

45. ⎟⎠

⎞⎜⎝

⎛−+= ∑

iihih

Chjjhjjhj PCCCTHPCCPCC min

,,min,, α

46. '''' servserv CGPCG =

47. ∑=i

ijj DIDIT ,

48. ITINVPC INVjjj γ=

V- International trade

49. ( )[ ] Xj

Xj

Xj

1

jXjj

Xj

Xjj D1EXTBXS ρρρ γγ −+=

50. jj

jXj

Xj

j DPL

PET1EXT

Xjσ

γγ

⎥⎥⎦

⎤

⎢⎢⎣

⎡ −=

51. ( )[ ] XRj

Xrj

XRj

1

ROWjXRjEUj

XRj

XRjj EX1EXBEXT ρρρ γγ '','', −+=

52. '','',

'','', ROWj

ROWj

EUjXRj

XRj

EUj EXPEPE1

EX

Xjσ

γγ

⎥⎥⎦

⎤

⎢⎢⎣

⎡ −=

53. 1jFOB

rj

rj0rjrj

Wj

Wj

PEPWE

EXDEXD −

⎥⎥⎦

⎤

⎢⎢⎣

⎡= σ

σ

θ,

,,,

54.

Wj

0j

0j

jj0jj XSTRADE

XSTRADEα

θθ⎥⎥⎦

⎤

⎢⎢⎣

⎡=

//

55. ( )[ ] Qj

Qj

Qj

1

jQjj

Qj

Qjj D1IMTBQ ρρρ γγ

−

−− −+=

56. jj

jQj

Qj

j DPMTPD

1IMT

Qjσ

γγ

⎥⎥⎦

⎤

⎢⎢⎣

⎡

−=

57. ( )[ ] MRj

MRj

MRj

1

RDMjMRjEUj

MRj

MRjj IM1IMBIMT ρρρ γγ

−

−− −+= '','',

58. '','',

'','', RDMj

EUj

RDMjMRj

MRj

EUj IMPMPM

1IM

MRjσ

γγ

⎥⎥⎦

⎤

⎢⎢⎣

⎡

−=

59. 0jj

0jjj PMTIMTPETEXTTRADE +='

39

60.

⎭⎬⎫

−−−−

⎩⎨⎧

+++=

∑∑

∑ ∑∑

Gr

Frh

h

Hrh

jrj

FOBrj

r

Rr

Rr

h

Rhr

jrjrj

TRReTRReTRReEXPEe

TRGTRFTRHIMPWMeCAB

,,,,

,,,

VI - Prices

61. jjjjjj DPDIMTPMTQPC +=

62. ∑=r

rjrjjj IMPMIMTPMT ,,

63. )()( ,,, jrjrjrj tx1tm1PWMePM ++=

64. )( jjj tx1PLPD +=

65. jjjjjj DPLEXTPETXSP +=

66. ∑=r

rjrjjj EXPEEXTPET ,,

67. FOBrjrj PEePE ,, =

68. ∑+=i

jiijjjj DIPCVAPVAXSP ,

69. agragragragragragragragr KDrkLATrdtLDTWTVAPVA ++= * redundant with EQ 4, 5, 6

70. nagnagnagnagnagnag KDrkLDTWTVAPVA += * redundant with EQ 6, 19

71. ∑=l

jljjj LDWLDTWT ,

72. agaadaladalagaagaagaaga LANrdagaWLANrdwLATrdt ,''''+=

73. agawaterawateraagaaialaialagaaga DIPCLANrdagaWLANrdw ,'''','''' +=

74. agapdalagppdalagpagpagpagp LANrdagpWLANrdwLATrdt ,'',''+=

75. agpwaterawateraagppialagppialagpagp DIPCLANrdagpWLANrdw ,'''','','' +=

VII – Labor market

76. ∑−=j

jllll LDLSLSU ,

77. MINll WW ≥

40

78. ( ) 0UWW lMIN

ll =−

VIII – Equilibrium

79. jjjh

hjj DITINVCGCQ +++= ∑ ,

80. CABSFSGSHITh

h +++= ∑

81. rjrj EXEXD ,, =

82. ∑=agr

agrlandSland LANLAN ,

83. ∑ ∑∑∑ ⎥⎦

⎤⎢⎣

⎡−+++=

j rrjrj

rrj

FOBrjjjjj

hhjj IMPWMeEXPEeINVPCCGPCCPCGDP ,,,,,

I- SECTORS All industries:

{

} SERVNMAN, WATERA, WATERNA, MINING,OMAN, TEXT, CHEM, IME, MCV,OAGRI, BEVER, SUGAR,CANNED,

OGR, OOIL, FLOUR, DAIRY, MEAT,FISH, OCROPS, INDCUL, LVST, VEG,OFRUITS, DAT, CITR, OLIV, LEGUM, OCER, BARLEY, HWHEAT, TWHEAT,Jji =∈,

Agricultural industries:

{}OCROPS INDCUL, VEG, OFRUITS,

DAT, CITR, OLIV, LEGUM, OCER, BARLEY, HWHEAT, TWHEAT,JAGRagr =⊂∈

Annual agricultural industries:

{}OCROPS

INDCUL, VEG, LEGUM, OCER, BARLEY, HWHEAT, TWHEAT,AGRAGAaga =⊂∈

Perennial agricultural industries: { }OFRUITS DAT, CITR, OLIV,JAGRagp =⊂∈ Other industries:

{

} SERVNMAN, WATERA, WATERNA, MINING,OMAN, TEXT, CHEM, IME, MCV,OAGRI, BEVER, SUGAR,

CANNED,OGR, OOIL, FLOUR, DAIRY, MEAT,FISH, LVST, NAGnag =∈

Labor skills:

{ } SWNAUWNA, SWA,UWA, FAW,Ll =∈

41

Land types:

{ }PDAL PIAL, ADAL, AIAL,LANDland =∈

Trading partner:

{ }ROW EU,Rr =∈

Households:

{ }URB RUR,Hh =∈

42

II- VARIABLES

jA : Total augmenting technical progress

LjA

: Labor augmenting technical progress

DagrA : Land augmenting technical progress

jbias : Labor technological bias

Dagrbias : Land technological bias

hjC , : Households h consumption of commodity j

min,hjC : Households h minimum consumption of commodity j

CAB : Current account balance

jCG : Public final consumption of commodity j

jCI : Aggregate intermediate consumption of sector j

hCTH : Household h consumption budget

jD : Commodity j produced locally

jiDI , : Intermediate consumption of commodity i by sector j

jDIT : Total intermediate demand for commodity j

hDIV : Dividend paid to household h

DTF : Firms direct taxes

hDTH : Household h direct taxes

e : Exchange rate

rjEX , : Export of commodity j to region r

rjEXD , : Export demand of commodity j to region r

jEXT : Total export of commodity j

G : Public expenditure

GDP : Gross domestic product

rjIM , : Imports of commodity j from region r

43

jIMT : Total import of commodity j

jINV : Investment in commodity j

IT : Total investment

jKD : Capital demand

agrlamdLAN , : Demand for land

SlLAN : Land supply

agrLAT : Demand for aggregate land bundle

jlLD , : Demand for labor

jLDT : Demand for aggregate labor bundle

lLS : Labor supply

jP : Producer price of commodity j

iPC : Composite price of commodity i

jPD : Consumer price of commodity j produced locally

rjPE , : Export price of commodity j to region r

FOBrjPE , : FOB export price of exports of commodity j to region r

jPET : Aggregated price of exports of commodity j

jPL : Producer price of commodity j produced locally

rjPM , : Import price of commodity j from region r

jPMT : Price of composite import of commodity j

jPVA : Value added price

rjPWM , : World price of commodity j imported from region r

rjPWE , : World price of commodity j exported to region r

jQ : Composite commodity j

agrrdt : Composite price for land in sector agr

landrdaga : Land price

agplandrdagp , : Land price

44

agrrdw : Composite price of irrigated land – water aggregate

jrk : Capital price

SF : Firms savings

SG : Government savings

hSH : Household h savings

jθ : Quality parameter

TI : Total indirect taxes

rTIM : Total tariff duties

jTRADE : Trade of sector j

GTRF : Transfers from firms to government

RrTRF : Transfers from firms to region r FTRG : Public transfers to firms

HhTRG : Public transfers to household h

RrTRG : Transfers from government to region r FhTRH : Transfers from household h to firms

RhrTRH , : Transfers from household h to region r

FrTRR : Transfers from region r to firms GrTRR : Transfers from region r to government H

rhTRR , : Transfers from region r to household h

lU : Unemployment rate

jVA : Value added of sector j

lW : Wages

agrWLAN : Demand for irrigated land – water aggregate

MINlW : Minimum wage

jWT : Wages

jXS : Aggregate output of sector j

45

hYDH : Household h disposable income

YF : Firms income

YG : Government income

hYH : Household h income

III- PARAMETERS

FA : Frontier TFP

VAjA : Scale parameter

jiaij , : Technical coefficient

Bjα : Bias parameter

BDjα : Bias parameter

Chj ,α : Marginal consumption of commodity j by household h

DHα : Land productivity-Human capital elasticity

DOPα : Land productivity-Openness parameter

Hα : TFP-Human capital parameter

1Hα : TFP-Human capital elasticity

2Hα : TFP-Human capital elasticity

OPα : TFP-Openness parameter

Wjα : Effect of trade on quality

jb : TFP-Human capital parameter

Djb : Land productivity-Human capital parameter

MRjB : Scale parameter (CES between imports by region)

QjB : Scale parameter (CES between IMT and D)

XjB : Scale parameter (CET between EXT and D)

XRjB : Scale parameter (CET between regions)

46

Ljβ : C-D Labor elasticity

Dagrβ : C-D Land elasticity

Kjβ : C-D Capital elasticity

jl ,γ : Repartition parameter

DIVhγ : Share of return to capital transferred to household h

DIVRrγ : Share of return to capital transferred to foreigners DWagrγ : Repartition parameter (CES between irrigated land and water)

INVjγ : Share of commodity j in total investment

LDagrγ : Repartition parameter (CES between land)

MRjγ : Share parameter (CES between imports by region)

Qjγ : Share parameter (CES between IMT and D)

Xjγ : Share parameter (CET between EXT and D)

XRjγ : Share parameter (CET between regions)

jio : Technical coefficient

Llh,λ : Share of wages from labor l received by household h

Dlandh,λ : Share of return to land received by household h

Khλ : Share of return to capital received by household h

hpms : Average propensity to save for household h

DWagrρ : Elasticity parameter (CES between irrigated land and water)

Ljρ : Elasticity parameter (CES between labor types)

LDagrρ : Elasticity parameter (CES between land)

MRjρ : Elasticity parameter (CES between imports by region)

Qjρ : Elasticity parameter (CES between IMT and D)

Xjρ : Elasticity parameter (CET between EXT and D)

XRjρ : Elasticity parameter (CET between regions)

DWagrσ : Elasticity (CES between irrigated land and water)

47

Ljσ : Elasticity (CES between labor types)

LDagrσ : Elasticity (CES between land)

MRjσ : Elasticity (CES between imports by region)

Qjσ : Elasticity (CES between IMT and D)

Xjσ : Elasticity (CET between EXT and D)

XRjσ : Elasticity (CET between regions)

Wjσ : Elasticity (World demand)

Ftd : Direct tax rate on firms income

Hhtd : Direct tax rate on households h income

jtm : Tariff rate on imports of commodity j

Ftr : Rate of transfers from firms to government

Hhtr : Rate of transfers from households h to government

jtx : Indirect tax rate on commodity j

jν : Technical coefficient

48

NESTED STRUCTURE OF PRODUCTION

OUTPUT XP

Leontief

Aggregate intermediate consumption (CI) Value Added (VA) CD

Intermediate demand by region of origin CES (Armington) Labor Land Capital NR CES CES

Skilled Unskilled Land demand by type

49

NESTED STRUCTURE OF CONSUMER DEMAN

DISPOSAL INCOME YD

Household savings SH Household expenditure on goods and services LES Armington demand by sector Aggregate imports by sector Aggregate domestic demand by sector Augmented CES Augmented CES Import demand by variety Domestic demand by variety CES Import demand by country of origin