Poverty and free trade agreement : the case of Jordan

(1)

U

NIVERSITEIT

VAN

A

MSTERDAM

Faculty of Economics and Business

Master Thesis

Poverty and Free Trade Agreement:

The Case of Jordan

Candidate: Matteo Ramina

ID number: 11088036

Supervisor: Dr. Prof. C. Elbers

Second reader: Dr. Prof. M. Pradhan

Academic year: 2016 / 2017

(2)

I

List of Figures and Tables

Figures:

Figure 1. Confidence on Free Trade Figure 2. Lerner Diagram for Jordan

Figure 3. Dynamics of the Lerner Diagram for Jordan Figure 4. The Heckscher-Ohlin Model

Figure 5. The Impact of Tariffs on Household Welfare Figure 6. Average Tariff Rate for the United States

Figure 7. Net Exports and Net Imports from the United States Figure 8. Breakdown of Exports by Product

Figure 9. Trends of the two groups without the free trade agreement Figure 10. Trends of the two groups with the free trade agreement Figure 11. Relationship between Expenditure and Tariffs

Tables:

Table 1. Jordan’s socio-economic indicators during the period 2000-2014 Table 2. Descriptive Statistics

Table 3. Descriptive Statistics (Breakdown between Rural and Urban) Table 4. First Stage Regression

Table 5. OLS and Second Stage Regression Table 6. Second Stage Regression (Breakdown)

2 5 5 6 11 17 17 18 25 25 34 16 33 34 35 36 37

(4)

III

Abstract

In April 2001, the Hashemite Kingdom of Jordan and the United States of America signed a free trade agreement that aimed at erasing tariff rates for the vast majority of tradable products with-in ten years. Nevertheless, with-in the period followwith-ing the implementation of the agreement, nation-al indicators of poverty raised unilaternation-ally. Hence, this thesis attempts to determine the causnation-al effect of the Jordan-US free trade agreement on the poverty of the Middle Eastern country, so as to assess whether the increase in the observed poverty can be explained by greater openness to international trade. This is accomplished by using a difference-in-differences methodology that is supported by an instrumental variable approach on repeated cross-sectional household survey data. The results indicate that 1 per cent raise in tariff rate causes a 0.1 percentage point in-crease in per capita consumption expenditure, which further rises to 0.3 percentage point for households living in urban areas or whose head is employed in the private sector. The effect is sizeable as it implies a decrease in consumption expenditure per capita of 2.18 percentage points and a consequent increase poverty headcount by 0.68 percentage points between 2002 and 2010.

Statement of Originality

This document is written by Matteo Ramina, who declares to take full responsibility for the con-tents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents

(5)

1

1. Introduction

In April 2001, the Hashemite Kingdom of Jordan and the United States of America signed a free trade agreement (FTA) that aimed at erasing tariff rates for the vast majority of tradable products within ten years. The agreement was the first of its kind between the United States and a Middle Eastern country (Bolle, 2001), and, besides causing a dramatic reduction in bilateral tariffs, it boosted trade between the two trading partners.

The new commercial relationship was one element of a broader economic reform undertaken by the Jordanian government, in which the opening of the domestic market to international trade was identified by the World Bank as a key factor to alleviate poverty (Worzala, 1994).

Nevertheless, in the years following the implementation of the agreement, the poorest segments of its population experienced a worsening of their living condition. National indicators of poverty raised unilaterally, with the headcount ratio increasing from 14.2 to 14.4 per cent between 2002 and 2010, while the poverty gap grew from 3.3 to 3.61_{. The effect was uneven between rural and}

urban areas, as the former reduced the share of people living in poverty while the former in-creased it.

Given the simultaneous presence of these phenomena, it is of interest to determine if the reduc-tion in tariffs caused by the FTA is responsible for the increase in the poverty observed in Jordan. The fact that bilateral and multilateral trade agreements have been a constant component of international relations for the last decades has spurred much research on the impact of greater openness to international trade on poverty. This is due to the fact that trade liberalisation un-leashes complex mechanisms whose causal effects are difficult to identify, since they are de-pendent on complementary policies, characteristics of factor markets and industries, socio-political context of the country, and details of the trade reform. At the same time, poverty is an intricate phenomenon, which is similarly affected by various elements like availability of infra-structure, policy responses, and household characteristics (Winters, McCulloch & McKay, 2004). The research community has contrasting opinions regarding this sensitive topic, and, as figure 1 illustrates, the viewpoint of the general public appears even more mixed. Figure 1 shows, for a selected number of countries, the percentage of people who believe that trade with other coun-tries leads to welfare gains thanks to an increase in salaries. It is interesting to observe that those countries experiencing the highest growth rates, which are also the poorest ones, believe that greater openness is associated with higher wages. On the contrary, people living in richer coun-tries have a much gloomier view, with a favourable view as low as 20 per cent of the time. Jordan lies in the middle, having a positive response rate equal to 45 per cent, showing neither a fierce opposition nor a strong support to free trade. This fact is likely to be related with a typical aspect of liberalisation policy, namely that such measure entails uneven distribution of benefits across the population, generating both winners and losers.

(6)

2 Figure 1. Confidence on Free Trade. Responses to the question “Does trade with other countries lead to an in-crease in the wages of (survey nationality) workers, a dein-crease in wages, or does it not make a difference?”. The average GDP growth rate is calculated during the period 2000-2014. Sources: Pew Global Attitudes & Trends (Spring 2014) and World Bank.

Hence, this thesis aims at determining the causal effect of the free trade agreement between Jordan and United States on the poverty of the Middle Eastern country, and it intends to identify who bore the cost of the agreement. Albeit divergent views have dominated since the results of the Heckscher-Ohlin model have been sternly challenged (Harrison, McLaren & McMillan, 2011), the prevalent opinion accentuates the beneficial consequences of greater openness. This study assesses if this has been the case for Jordan as well.

Some scholars criticise this kind of endeavour, as they deem futile to isolate the causal impact of trade policies on poverty. According to the critics, the numerous implications of trade liberalisa-tion on the whole society, and complementary reforms that are oft undertaken during the same period, undermine any empirical research. This can be especially true for a country like Jordan, which has implemented a political agenda that featured several reforms during the last five dec-ades (Yousef, 2004). Therefore, an appropriate identification strategy needs to be designed so as to disentangle the numerous elements that might have conditioned poverty and to identify the contribution of the FTA.

Because of the impracticalities to perform a study which encompasses all possible effects origi-nating from the agreement, I restrict the analysis to the short-term impact directly caused by the FTA. More precisely, my focus is on those channels that deal with variations in prices of produc-tion factors and changes in the labour market. These are mainly determined by households’ re-sponses to changes in wages and to employment opportunities. The choice of disregarding

oth-BGD BRA GBR CHL CHN COL EGY _SLV FRA DEU GHA GRC IND IDN ISR ITA JPN JOR KEN LBN MYS MEX NIC NGA PAK PER _PHL POL _RUS ZAF _KOR SEN ESP TZA THA TUN TUR UGA UKR USA VEN VNM 0 20 40 60 80 100 -2 0 2 4 6 8 10 'In cr ea se ' r es po ns es (p er c en t)

(7)

3 er channels does not suggest that they are irrelevant for the poverty analysis, but this restriction assures a better identification of the relationship under scrutiny. The econometric analysis is in-spired by Topalova (2007), who employs a difference-in-differences method that is supported by an instrumental variable approach. However, due to the use of different datasets an adaptation of her methodology is required, which is explained in section 4. Furthermore, the analysis is de-composed with respect to geographic and demographic features of the household.

The study is performed by using repeated cross-sectional household survey data on consump-tion expenditures and informaconsump-tion on bilateral tariff rates between Jordan and the US. Jordan’s Department of Statistics (DoS) designed and collected the Expenditure and Income Household Surveys (HEIS) in 1992, 1997, 2002, 2006, 2008, and 2010. The HEIS are disseminated in a limited way by the Economic Research Forum (ERF), which was also the provider for the present study. The surveys do not merely supply a large amount of information on household consumption be-haviour, but provide other relevant information such as the sector of employment of the house-hold’s head, a fundamental element of the model. Tariff rates are extracted from the Trade Anal-ysis Information System (TRAINS) constructed by the United Nations Conference on Trade and Development (UNCTAD) and disseminated through the World Integrated Trade Solution (WITS) by the World Bank. Finally, additional data, such as the number of people employed per sector, is directly obtained by the Department of Statistics of Jordan.

The remainder of the thesis is the following. Section 2 presents the literature review, firstly illus-trating the channels through which greater openness to international trade impacts poverty, and secondly by reviewing the empirical research. Section 3 describes the liberalisation policies that Jordan has undertaken and the FTA with the United States. Section 4 explains the methodology employed to construct the poverty profile and to determine the causal effect of the FTA on pov-erty. Section 5 provides a description of the datasets and the variables used in the analysis. Sec-tion 6 presents the results, which are further discussed in secSec-tion 7. Finally, secSec-tion 8 concludes.

(8)

4

2. Theories of International Trade, Poverty and Empirical Literature

2.1. The Neoclassical Theory

One of the main theories that can be used to disentangle the relationship between free trade agreements and poverty is the Heckscher-Ohlin model (HO model), one of the pillars of interna-tional trade theory in neoclassical economics. Although new theories and empirical evidence challenges its conclusions, the model is still widely used for academic scopes (Leamer, 1995), and it has been used to encourage liberal trade policies since the post-war period (Harrison, McLaren & McMillan, 2011). Indeed, the HO model is deemed to give important and concise information in explaining the gains from trade (Krugman, 1990), while it is at the root of some principles of the World Trade Organization (Smit, 2010).

The HO model was developed by the economists Eli Heckscher and Bertil Ohlin during the 1920s, and it is arguably amongst the most important advancement in trade theory that followed the Ricardian model of comparative advantage. Its usefulness stems from the illustration of the neoclassical view on the production behaviour of countries trading with each other. Specifically, the model draws conclusions on the specialisation patterns of production and, by looking at the changes on the returns of production factors, it gives the redistributive outcome resulting from the transition of an autarkic regime to one of free trade. The following paragraphs describe the model and outline its most important conclusions.

The basic framework is characterised by two countries that manufacture identical goods by means of the same production process. The latter is featured by constant returns to scale, and economic agents operate within a perfect, competitive market. Furthermore, countries possess equal homothetic preferences, while resources are distributed unequally across countries and endowments are fixed and cannot be transferred (but they can be reallocated across the pro-cesses of production). In order to give a description in harmony with the present study, assume that the two countries are Jordan and the United States, which manufacture textile and dairy products. Additionally, suppose that the two goods are produced by using only wage labour and capital.

After having outlined the assumptions of the HO model, the analysis begins by focusing on the production choice of Jordan. This one can be determined thanks to the Lerner Diagram (figure 2), which shows a hypothetical isoquant map for the manufacturing of textile and dairy with the respective isocost line. For simplicity, the analysis is undertaken considering unit-value isoquant curves,2_{meaning that the graph shows the combinations of labour and capital that produce one}

2_{An isoquant map is a diagram comparing isoquant curves. An isoquant curve shows the set of points that}

per-mits the production of one unit of output while changing the combination of the production factors. The isoquant curve is an important element in determining the production choice of an economic agent – in this ex-ample Jordan, – as it solves for the cost-minimisation problem of the two products to be manufactured. The fact that unit-value isoquant curves are examinedimplies that only the combinations of production factors that yield one unit-value of output are considered. The two unit-value isoquant curves are connected by the isocost line, namely the set of points that allows to produce the same amount of output using different combination of pro-duction factors. When this occurs, equilibrium is reached and profits are equal to zero. The current

(9)

representa-5 unit-value of output. In this production process capital is rented at the rate r, and labour is paid at the prevailing wage w. Of particular interest are the tangency points 2 and 1, in which the re-spective output of textiles and dairy is maximised for 1 monetary unit of input. In such loci w1 < w2 and that r1 > r2, implying that the production of dairy is relatively capital intensive, whereas textile is manufactured using relatively more labour. At this stage an autarkic regime is assumed, thus Jordan produces both goods regardless of the relative abundance of the factors of produc-tion.

Before observing the effects of opening the market to international trade, the Lerner Diagram

can be used to demonstrate one of the fundamental tools of the HO model, i.e. the Stolper-Samuelson theorem (SS theorem). Assume that a sudden increase in the price of textile occurs. This implies that the value of output produced per fixed amounts of production inputs is higher, which causes the unit-value isoquant for textiles to shift towards the origin of the axes (figure 3). At this stage, unexploited profits are available in the textile sector, since the previous isocost curve (depicted in figure 3 by the lighter straight line) intersects the new unit-value isoquant curve. This causes the isocost line to rotate clockwise until it reaches the new equilibrium in point 3, reshuffling the allocation of the inputs of production into the textile sector. In such point w1 < w3 and r1 > r3, entailing that the wage rises at the expenses of the rental rate of capital. Conse-quently, the price increase of textile determines a rise in the returns from labour, i.e. the input used most intensively in relative terms by such sector. Hence, this exemplifies the general con-clusion of the SS theorem, that is an increase in the relative price of a product induces a rise in the rate of return of the production factor which is used most intensively in relative terms for the tion of the production process implies constant returns to scale, as this enables to represent the production func-tion can be represented by any isoquant. For a detailed theory of about the Lerner Diagram see Leamer (1995.) Figure 2. Lerner Diagram for Jordan. The straight line

represents the isocost curve, whereas the dashed lines are the isoquant curves for textile and dairy.

Figure 3. Dynamics of the Lerner Diagram for Jordan. The lighter straight line indicates the position of curves in Figure 2. 1 / r2 2 1 / w1 Dairy Textile 1 1 / r1 1 / w2 Ca pi ta l Labour 1 / r2 1 / w2 Dairy Textile 1 / w3 1 1 / r3 1 / r1 1 / w1 2 3 Ca pi ta l Labour

(10)

6 manufacturing of such good. Wherefore, the rate of return of the other factor is bound to de-crease, due to a reduction in the relative price of the good that uses it the most intensively. The SS theorem is a fundamental toehold in order to study the HO model. Figure 4 combines the production possibility frontiers of both Jordan and the US3_{, alongside with their indifference}

curves with and without an autarkic regime. When the two countries have a closed economy their relative demand curves are equal, due to their identical, homothetic preferences. However, the differences in their relative abundance of resources cause unequal production possibilities fron-tiers, which are skewed towards the manufacturing of the product which uses intensively the fac-tor of production relatively more abundant. Also, different endowments entail unequal relative price of inputs of production, depicted by the different steepness of the lighter orange straight lines. In this example, labour is relatively more abundant in Jordan than in the US, while the con-verse holds true for capital. Therefore, Jordan’s production possibility frontier is skewed towards the textile sector and intersects its indifference curve in point 1. At the same time, US production

possibility frontier is skewed towards the dairy sector and meets its indifference curve in point 2. When the countries dismantle their trade barriers, each economy does not have to consume all of its production anymore, rather it can specialise in the manufacturing of the products for which it has a comparative advantage with respect to its trading partner. In this new context, Jordan concentrates on the production of textile, whereas the US on the production of dairy. Moreover,

3_{The production possibility frontier is concave, meaning that the products are imperfect substitute.}

Figure 4: Heckscher-Ohlin Model. Indifference curves are depicted by the convex curves, while the production possibility frontiers are depicted by the concave ones. The budget line is given by the straight lines.

i3 i2 i1 2 ₃ 1 Te xt ile s Dairy

(11)

7 another consequence of opening up the domestic market is that the price differences are erased due to the new production choices. Indeed, for a given good there will be relative excess supply in one country and a relative excess demand in the other country, so that this good will be ex-ported by the first country and imex-ported by the other one.

Recall that, under autarky, consumption equals production and the equilibrium is reached where the indifference curve is tangent to the production possibility frontier. With openness to interna-tionally commerce, countries are not constrained to consume all of their production, and they can trade at any point along their budget constraint. In this way, the relative price of the good that intensively uses the relatively abundant factor rises, while the relative price of the good that intensively uses the relatively scarce factor falls. Thus, Jordan shall export textile and US shall ex-port dairy products, achieving the new equilibrium in point 3.

With respect to the poor in developing countries, the neoclassical narrative emphasises that the short run benefits are twofold. Firstly, when the two countries embrace free trade, American and Jordanian consumers can attain indifference curves of higher level, which is represented in figure 4 by the fact that i3 > i1 and i3 > i2. Obviously, this result does not discriminate according to the percentile of the income distribution where the household lies, and it affects positively all con-sumers. The poor can reap the benefits of the trade liberalisation thanks to the distributional changes determined by the adjustments in the relative prices of textile and dairy, which adapt to a new world price. More precisely, assuming that wage earners (who are used intensively in the production of textile) are poorer than capital owners, an export-led increase in demand for tex-tile boosts the salaries paid in that sector. Thus, the wage of Jordanian workers rises, causing a reduction of poverty and the improvement in their standard of living.

2.2. New Trade Models

The result just analysed provides a powerful theoretical foundation to promote free trade poli-cies in developing countries. Nevertheless, both the standard HO model and the SS theorem have been mauled for many decades, particularly since the beginning of 1990 (Harrison, McLaren & McMillan, 2011). The appropriation of the gains from trade expected by the neoclassical view are claimed to be based on a theory that is deficient in several aspects. Firstly, the standard HO model is built on too many assumptions exploiting homogeneity between the two trading part-ners that are oft not necessarily so similar. For instance, the framework does not consider heter-ogeneous firms, lacking the effect that firm-specific productivity exerts on export choices. Sec-ondly, the HO model does not consider technology transfer and is not adequate to inspect the impact of progressive technology advancements (Krugman, 1979). The fact of neglecting tech-nology transfer in the aftermath of a trade liberalisation is an important flaw, as skill-biased ex-ports might prevent unskilled workers to benefit from the reform (Winters et al., 2004). Thirdly, global-value-chains and intra-industry international trade are disregarded, posing a threat in its ability to explain current patterns of commerce (Schiff & Winters, 1998). Fourthly, the model ex-amines homogenous products only, and this is not sustained by the empirical evidence that highlights the role played by product differentiation on the production choice. Fifthly, the model

(12)

8 does not allow for movements of the production factors across countries, ignoring the cross-country movements of people and capital and their impact on factor endowments (Harrison, McLaren & McMillan, 2011).

Academics have not been indifferent to these problems, so that a number of improvements on many aspects aforementioned have been accomplished4_{. Nevertheless, Harrison et al. (2011)} maintain that the HO model still falls short of explaining the lack of poverty alleviation and the raising inequalities that have been observed in several instances of trade liberalisation (see sec-tion 2.3 for the review of the empirical literature). Therefore, a parallel strand of literature that departs from many assumptions of the neoclassical framework has developed.

The most important theory diverging from the neoclassical narrative is probably the model costructed by Melitz (2003), which is an international trade model featured by firm heterogeneity and monopolistic competition. The implementation of these elements into an international trade model brings about several consequences, amongst which the fact that trade liberalisation does not cause a unilateral, positive effect on the poor of developing countries. In order to see how such conclusion is achieved, the Melitz model is explained in the following paragraphs.

Assume that an economy is characterised by consumers that take part in the production process thanks to the possibility of establishing a firm and become an entrepreneur5_{. In the event one}

consumer decides to do so, he or she faces the following production function: 𝑞_",$= (𝑙_",$− 𝑓)𝜙_"

This equation determines the output 𝑞 produced by the firm 𝑖 at time 𝑡, who is using labour 𝑙 as the only means of production, while incurring a constant fixed cost 𝑓 common to all agents in-volved in production. Heterogeneity among firms is determined by the parameter 𝜙, which is the unknown productivity factor of the entrepreneur. This one varies across producers and is un-veiled only once the entrepreneurs sets up his own business activity. A closed economy charac-terised by such production function achieves its equilibrium at a cut-off value 𝜙._{= 𝑓}_{, where firms}

that possess a 𝜙"< 𝜙. go out of business, whereas those whose productivity is 𝜙" > 𝜙. carry on

their activity.

This specification produces striking results when considering the new equilibrium condition orig-inating from the transition to an open economy. Suppose that, in a regime of free trade, a firm willing to export faces a transport cost equal to 𝑐, which is also constant across firms and time. In order to generates higher profits, firms have the incentive to sell their product in the foreign market. However, since exporting implies a total cost made up of 𝑐 and 𝑓, only more productive firms have the capability to meet such higher cost. Consequently, a new cut-off value 𝜙2_is

estab-lished at equilibrium, which is higher than 𝜙._{, that is 𝜙}2_{> 𝜙}._{. Although Melitz has not}

devel-oped its model to observe the consequences on poverty, many economists have taken

4_{For an exhaustive source on this topic see Feenstra (2015), who explains some of trade models that deal with a}

number of the aforementioned issues, like Feenstra and Hanson (1996) and Davidson, Martin and Matusz (1999).

(13)

9 vantage of this new equilibrium condition and formulated trade models that relate greater openness to a worsening of the welfare of poorer households.

A notable work is the one by Egger and Kreockemeier (2009), who integrate frictions in the la-bour market into the Melitz model. Specifically, the authors make usage of the efficiency-wage hypothesis to assume that employers pay employees higher than the market-clearing wage. This does not merely allow for equilibrium unemployment, but also for cross-firm differences in the wages paid to employees, hence generating inequality which is exacerbated by trade. Indeed, the rise in the cut-off equilibrium value of 𝜙 has two detrimental effects on the poor. Firstly, to the extent that the marginal increase in productivity outnumbers the marginal increase in output, unemployment increases because of a lower demand for labour. Secondly, trade drives a wedge between the wage paid by high productive firms and inefficient ones, making the poorest em-ployees worse off. Additional work on this topic is done by David and Harrigan (2011) and Help-man, Itskhoki and Redding (2010), who formulate models using different types of labour market frictions which both come to the conclusion that greater involvement in international trade is not associated to a lifting of people out of poverty. Rather, free trade can cause severe hardship to the poor in developing countries and should thus be discarded from the key measures to allevi-ate poverty.

2.3. Empirical Literature

In spite of the theoretical beauty of both approaches, they scarcely transpose their results into reality, and they do not permit to fully understand the complex phenomena observed across decades of trade reforms. As a matter of fact, the Heckscher-Ohlin model and the Stolper-Samuelson theorem encountered numerous waves of studies challenging their validity with real-world data. In the 1960s and in the 1970s, researchers demonstrated that the neoclassical ap-proach could not explain the intra-industry trade observed at that time (Baldwin, 1971, Neary, 2004). Later on, during the 1990s, failure of the HO model and the SS theorem to predict the change in the income distribution of many Latin America economies undermined its legitimacy (Harrison et al., 2011). These led Davis and Mishra (2007) to affirm that “The Stolper-Samuelson theorem is dead” and that “A starting point [for trade theory, e.d.] is to cast off the shibboleths of Stolper-Samuelson in its global form as a useful way to think about the world that we actually live in”.

Nonetheless, Harrison et al. (2011) admits that sound evidence of an adverse effect of greater openness on poverty is scarce, as only a few studies located in India and Mexico permit to infer a negative causality. Contrarily, the Four Asian Tigers6_{provide a typical example where poverty}

reduction went hand in hand with trade liberalisation. Therefore, as maintained by Winters (2002), theory can provide only a partial explanation, and the final answer on the relationship be-tween free trade and poverty is of empirical matter.

(14)

10 A large quantity of empirical studies is grounded on the framework constructed by Winters, McMulloch and McKay (2004), who inspect the effect of trade liberalisation on the welfare of poor households by looking at four main channels. The first one has a macroeconomic perspec-tive and relates the effect of deeper integration with global markets to economic growth and macroeconomic stability. The second channel is given by the responses of households to price changes of marketed goods, where poverty is affected by the adjustments in relative prices of the products that the household consumes or produces. The third element is given by the movements in the labour market, in which a rise or a decline in the prices of goods determines a corresponding increase or decrease in the returns to wage labour. Finally, poverty may be af-fected by lower government support through social safety nets and other schemes, as public fi-nances may be curtailed by the dismantling of tariff barriers.

Thanks to the intense research on the topic, surveys of the literature by Goldberg and Pavcnik (2004), and later Winters and Martuscelli (2014), corroborate that the effects of greater openness on the poor are well known and foreseeable, since common responses to the reforms have been

Figure 5. The Impact of Tariffs on Household Welfare. Adapted from Winters (2002). Household consumption – which is the focus of this diagram – is directly affected by three main channels. Economic growth – the fourth element identified by Winters – does not cause a direct micro-stimulation to market prices or government trans-fers, consequently it does not appear in this figure. Dashed lines indicate that, in the econometric model used in this thesis, the element is not directly affected by household consumption. The literature on market prices is ana-lysed insofar as they affect household consumption indirectly through factor markets.

(15)

11 observed. Notwithstanding this reassuring fact, the same authors affirm that the outcome deriv-ing from one specific trade reform is hard to predict, due to the high complexity that imbues the channels of transmission and the originality of the contexts.

As mentioned previously, four main channels transmit economic shocks at the border to poor households. Figure 5, which is an adapted version of the diagram presented in Winters (2002), illustrates graphically the elements involved. These operate through various mechanisms, which differ for, inter alia, timing of the impact, demographic factors affecting household welfare, and average effect on poverty. Moreover, they exert contrasting forces on poverty. This hinders the pursuit of a study that encompasses all channels of transmission, and gives priority to analyses that disentangle the various mechanisms and that focus on specific links. So, despite the contri-bution of all channels is fundamental in understanding the complete picture, a narrower focus is desirable so as to prove causal inference. Thus, the following sections present the empirical evi-dence on the channel inspected in this work, namely the changes in the labour market. Further-more, a brief review is given on the responses of households and goods markets, due to the at-tention that this channel has gathered in the literature.

2.3.1. Labour Market Channel

As previously explained, the Heckscher-Ohlin model predicts that, following a trade liberalisa-tion, countries which abound of unskilled workers are expected to benefit from an increase in the demand for such unskilled labour. Consequently, the wage of those workers (who are likely to belong to poor household) would increase. However, these benefits may not materialise if there are frictions in the labour market. Similarly, empirical evidence suggests that free trade influ-ences poverty either way.

Proof of positive effects is given by Milner and Wright (1998), who show that trade liberalisation raised wages of the clothing industry in Mauritius, where the majority of workers are women. Fur-ther evidence of a positive link between trade and poverty reduction is provided by Jerkins (2004), who inspects the changes in the labour market following the trade reform implemented in Vietnam during the 1990s. Specifically, the author looks at the impact of greater integration to global markets on employment growth and finds a positive relationship on the creation of un-skilled jobs. Indeed, trade liberalisation boosted exports for labour intensive industries like gar-ments and footwear manufacturing, and this contributed positively on the alleviation of poverty. Other two studies by Heo and Doanh (2009) and McCaig (2011) examine the same trade liberali-sation in Vietnam and obtain comparable results. Heo and Doanh (2009) maintain that the growth in real wages observed in the agricultural and manufacturing sectors played an important role in explaining the reduction of poverty observed in rural areas. McCaig (2011) finds that rural workers experienced an increase in the wage premiums within province, especially for those em-ployed in agriculture, forestry and fishing. Moreover, the author detects a dynamic labour market which quickly reallocated workers from the aforementioned sectors into manufacturing, thanks also to leading export industries that fostered job creation. Consequently, McCaig’s

(16)

economet-12 ric results determine that an increase in provincial exposure to trade by one standard deviation caused a reduction in the poverty headcount of that province by 10 per cent.

However, many empirical studies conducted in Latin America, India and other regions obtained opposite findings, raising concerns that greater involvement in international commerce might bring on skill-biased demand for labour, which would have detrimental effects on the poor (Atta-nasio, GoldBerg & Pavcnik, 2004). For instance, Hanson and Harrison (1999) inspect the differ-ences in wages and employment opportunities between skilled and unskilled workers in the af-termath of the trade liberalisation in Mexico. The authors find that the dismantling of trade barri-ers affected disproportionately unskilled workbarri-ers, who were employed in industries that were highly protected prior to the reform. Therefore, although they do not identify great changes in the employment shares between the two categories, they do observe a widening in the skilled-unskilled wage gap. This conclusion is corroborated by Nicita (2004), who provides further evi-dence that the trade liberalisation in Mexico created benefits there were relatively larger for the skilled labour than the unskilled labour.

Moreover, a survey of the literature by Santos-Paulino (2012) concludes that, despite trade policy is industry-specific, variation on wages is likely to be firm-specific. This is due to the fact that only employees who work in a firm with high impact on export may benefit from trade liberalisation, and, to the extent that few firms benefit from the greater openness, the impact on the overall wages may be small. The importance of firm characteristics is also highlighted by Currie and Har-rison (1997), who provide evidence for an Arab country by examining the liberalisation of trade in Morocco. The authors find that firms which excel in exporting were characterised by a larger use of temporary workers and by a lower disbursement of payrolls. Lower wages did not change after the reform, so that the trade liberalisation failed to increase the salary of unskilled workers. In addition, Currie and Harrison observe that many elements contribute to the protection of do-mestic wages and keep them inefficiently high. Since trade liberalisation entails the elimination of these elements, or at least their shrinkage, it is a possible source of wage reduction. Similar findings are obtained by Mouelhi (2007), who performs a research on the effect of trade liberali-sation on the demand skill of firms in Tunisia.

Another strand of literature concentrates on labour market frictions and how they prevent the reallocation of workers across industries and regions. Topalova (2007, 2010) provides important findings on this topic, by observing the effect of the trade liberalisation in India. She finds that rural districts which experienced a larger decrease in tariffs were associated with lower reduction in poverty rates compared to rural districts that maintained higher tariffs. Topalova identifies the immobility of the labour force across regions as the major determinant behind this result. In-deed, the impossibility for rural households to migrate towards those districts that were positive-ly affected by the trade liberalisation hindered the benefits of greater openness. Although not focusing on poverty, similar evidence is observed after the 1990s trade reform in Brazil by Menezes-Filho and Muendler (2011). The authors find that cuts in tariff rates were followed by a slow reallocation of workers from inefficient firms to more productive ones, while decreasing hir-ing rates.

(17)

13 Finally, an interesting study is provided by Castilho, Menéndez and Sztulman (2012), who also inspect the trade liberalisation in Brazil and find that, similar to Topalova (2007, 2010), areas more exposed to the reform experienced lower poverty reduction. However, contrarily to Topalova, urban agglomerates were the most adversely affected areas, so that a one percentage point in-crease in their trade liberalisation index raised poverty headcount ratio by 0.67 percentage points. At the same time, Castilho, Menéndez and Sztulman could not find a similar effect for rural areas, whose poverty level was not correlated with greater openness.

2.3.2. Households and Good Markets

Shocks in imports and exports lead to variations in the prices of goods and services. Those changes have a direct effect on poor households, as they could either consume or employ in their production function such goods and services. The final impact on the poor’s welfare de-pends upon their economic position, i.e. whether they are net producers or net consumers. Some studies on the liberalisation process that has taken place in Vietnam during the last dec-ades have found an average positive effect, like for instance Niimi, Dutta and Winters (2007). By focusing on the rice sector the authors find that, on average, the consumer benefit from import-ed goods outweigh the loss of producers, whereas the producer benefit from more expensive export goods outweigh consumer loss. However, despite the overall positive impact on poverty, several studies show how this impact is conditioned on several aspects. The improvement in liv-ing standards is uneven across the poor, who can often reap the benefits if they fulfil specific re-quirements. For instance, two other researches on the same trade reform in Vietnam (Glewwe, Gragnolati & Zaman, 2002, and Justino, Litchfield & Pham, 2008) find that the beneficial effects of trade liberalisation were conditioned on location and ethnicity of the household, education, oc-cupation and sector of employment of the household’s head, and also geographical aspects.

2.4. Measuring Trade Openness

The measurement of a country’s openness is a challenge at the heart of the present study, and the empirical literature has used a vast number of trade-related variables. These are divided by Spilimbergo, Londoño and Székely (1999) into two main groups. The first one includes the so-called incidence-based variables and comprise mainly tariffs and proxies for non-tariffs barriers and trade policy, whereas the second group includes outcome-based variables like export to GDP ratio, trade volumes, and foreign direct investments. Justino, Litchfield and Pham (2008) claim that the latter group performs poorly in explaining concrete changes in household living standards, while Goldberg and Pavcnik (2004) affirm that tariff changes are a perfect candidate to identify the effect of the trade reform on poverty, when the tariff change is the main element of the trade liberalisation.

Although tariffs represent a main hurdle for countries willing to export, many other elements contribute to hinder international trade. One of the main problems regards taking into account nontrade barriers (NTBs), whose non-quantitative nature complicates their inclusion in the

(18)

analy-14 sis. NTBs encompass a large array of elements that hinder activities of import and export, such as quantitative restrictions, licensing requirements and regulations which differs from those re-quired by domestic firms. Not only the measurement of NTBs is a complex issue, but its compar-ison across time is difficult too. This problem is especially relevant for developing countries (Goldberg and Pavcnik, 2004), and when considering single policy changes like a free trade agreement.

The question is less worrisome if the NTBs are positively correlated with tariffs, since, despite the upward bias one would estimate, the coefficient would still provide relevant information. Sound evidence on this topic is available for certain countries, but it is not for Jordan. Tudela-Marco, Garcia-Alvarez-Coque, and Martinez-Gomez (2014) provide the only analysis on the relationship between NTBs and tariffs in Jordan, without finding any significant relationship. The lack of inclu-sion of NTBs is a factor that weakens the econometric results of the present study. However, the absence of specific data on NTBs related to the Jordan-US FTA prevents the addition of this measure.

(19)

15

3. Liberalisation Policies in Jordan

3.1. A Path-Breaking Reformer

Yousef (2004) documents that, similarly to other Middle Eastern countries, Jordan was historically ruled by an interventionist government that used to implement generous redistributive measures. Jordan was also a rather closed economy, that was protecting its economy with high tariff rates. This caused, as the author says, the failure in attracting the economic interests of in-dustrialised countries in the awake of globalisation. Amerah (1996) singles out two objectives that explain such restrictive policy, namely the promotion of an import substitution industrialisation, and the funding of public financing for social expenditure. Besides the goals of the Jordanian government, regional instability was one of the main reasons for such isolation. This one did not merely hinder the integration of Jordan into global markets, but it also severely hit its own econ-omy. For instance, the 1990 Persian Gulf War caused massive expatriation of hundreds of thou-sands of Jordanian migrants back to Jordan, putting pressure on the labour market that had to cope with a sudden influx of workers and a reduction in the amount of remittances.

In the 1980s, a combination of adverse factors spurred Jordan to reform profoundly its economy. Oil prices plummeted, and this, although Jordan is not an oil-producer country, contributed to the deterioration of the public finances that were already heavily constrained by the large redis-tributive measure (Amerah, 1996). At the same time, total factor productivity was stagnant and the rate of capital accumulation slackened (Yousef, 2004).

Therefore, in order to cope with the worsening of the macroeconomic conditions, Jordan em-barked on market-oriented and growth-enabling economic reforms, becoming path-breaking reformer amongst the Arab countries (Dasgupta, Keller & Srinivasan, 2005). The pursuit of liber-alise trade was a key element of this agenda, which gained credibility in the early 1990s thanks to the expression of interest to begin the accession process to the World Trade Organization (Amerah, 1996).

The initial measures entailed a reduction of nominal tariffs and a revision of the investment law, but, albeit a decline in some nominal tariffs was recorded, Amerah finds that the early attempts to dismantle the barriers to trade achieved mediocre results.

Efforts to reform trade policy rejuvenated in 1997 (Al Khouri, 2000), when Jordan passed legisla-tion that aimed at simplifying the tariff landscape by consolidating monetary burdens other than import duties into tariffs7_{. Nonetheless, the new measures were unsatisfactory in many respects.}

For instance, several fees were actually not included into the law, so that the attempted consoli-dation was only partial and confusing. Also, many of these taxes and duties would differ

7 Duties and Taxes Consolidation Law No. 7 (1997). The other fees and taxes that were not absorbed in the tariff rates where the general sales tax (GST), the Sales Surtax and an export surcharge of 5% imposed under Import and Export Regulation No. 74 of 1993.

(20)

16 ing to many aspects of the product, such as the country where it was manufactured8_{, the nature}

of the good9_{, the different variations of the same product}10_{, the stage in the global-value-chain}11

and legal entity of the importer12_{. Therefore, Jordan entered the new millennium with much to}

do on trade liberalisation.

Nevertheless, a complementary measure issued by the Ministry of Industry and Trade in 199713

attempted to lessen nontariff barriers, by reducing mandatory license requirements from all im-ports. At the same time, Jordan initiated a series of bilateral and multilateral agreements, the first one being the Qualifying Industrial Zone (QIZ), stipulated with the United States in 1996. The QIZ guarantees duty-free access for goods manufactured in Jordan to the American, as long as they include inputs from Israel.14_{Two years later, in 1998, Jordan encouraged the revival of the}

Agreement for Facilitation and Promotion of Trade, an agreement in effect with the members of the Arab League which aimed at facilitating trade amongst the participating countries. Notwith-standing these developments, Busse and Gröning (2012) maintain that tariff and nontariff barriers were still significant and were lagging behind the planned elimination schedule.

Intentions to join the WTO materialised in April 2000, when Jordan finalised the accession pro-cess, and in October of the same year it signed a free trade agreement with the US. The latter agreement was the first of its kind with a western country and it is carefully analysed in section 3.2. Beside the FTA with the US, Jordan signed several other free trade arrangements in the first decade of the 2000s, amongst which with the European Union, Canada and Singapore (Busse & Gröning, 2012).

The liberalisation of trade led the average tariff rate to decrease by almost 10 percentage points between 2000 and 2014, declining from 18.9 to 9.3 per cent (table 1). At the same time, trade statistics signal a larger role played by international commerce in the economy, with imports in-creasing by 68.5 to 87.9 per cent of GDP and exports by 41.8 to 53.9 per cent of GDP during the same period.

8 North African and Middle Eastern countries would have a favoured regime, but they would still have these fees applied, which would still differ from country to country.

9 The rate of GST would differ for air conditioning units, marble, cameras, microwaves, and several other prod-ucts.

10 Cars would bear different burdens according to, for instance, their engine size. 11 Goods in transit are exempted from GST.

12 Mosques, churches and orphanages would not have to pay GST. 13 Ministry of Industry and Trade’s Regulation No.1 1997

14 The QIZ was basically an extension of the Israel-US free trade agreement that expanded the range of goods covered by the policy.

(21)

17 Variable 2000 2002 2006 2008 2010 2012 2014 GDP (million) 36.9 38.9 41.1 42.8 46.5 50.3 54.4 GDP growth rate 4.2 5.3 5.8 4.2 8.6 8.2 8.1 GDP per capita 7,742.5 8,011.6 8,323.4 8,491.8 8,988.8 9,428.5 9,828.1 Population (million) 4.8 4.9 4.9 5.0 5.2 5.3 5.5 Rural population 20.2 19.9 19.6 19.4 19.1 18.8 18.6 Poverty headcount ratio - 14.2 13.0 13.3 14.4 - - Rural poverty headcount ratio - 18.7 18.7 - 16.8 - - Urban poverty headcount ratio - 12.9 11.8 - 13.9 - - Poverty gap - 3.3 2.8 2.6 3.6 - - Employment ratio 36.2 35.3 34.8 34.7 35.6 35.0 35.2 Unemployment rate, total 13.7 15.8 16.2 15.4 12.4 14.8 14.0 Unemployment rate, youth 27.9 33.6 34.8 34.4 27.4 32.5 30.3 Imports (per cent of GDP) 68.5 67.2 66.6 68.4 82.5 94.2 87.9 Exports (per cent of GDP) 41.8 42.1 47.4 47.4 52.2 52.7 53.9 Average tariff rate 18.9 12.1 12.7 11.4 11.2 12.0 9.3

Table 1. Jordan’s socio-economic indicators during the period 2000-2014. All monetary values are given in con-stant 2011 international $ computed through the PPP method, except for the values of imports and exports, giv-en by constant 2005 US$. Poverty headcount ratios and the poverty gap are measured in percgiv-entage terms and are determined by national poverty lines, which are 392 Jordan Dinar (JD) in 2002, JD556 in 2006, JD680 in 2008, and JD814 in 2010. Source: World Bank.

3.2. The Jordan-US Free Trade Agreement

Because of the refusal of Jordan to join the US-led coalition against Iraq during the Persian Gulf War, the US Congress did not undertake any large-scale initiative to assist the Jordanian econo-my prior to 1995. However, soon after the peace treaty on 26 October 1994, a number of initia-tives were taken, which included the boost of bilateral economic and military assistance, debt forgiveness and the establishment of the aforementioned QIZ. As a matter of fact, US aspiration to make a fresh start with Jordan were grounded more on political than economic ground. In-deed, Rosen (2004) maintains that “The US-Israel and US-Jordan FTAs are clear examples of the use of trade policy – specifically bilateral free trade agreements – as a means of pursuing foreign policy objectives”. Rosen affirms that the American geopolitical stake in the region outweighs economic interests, so that the main reason for the US to implement this agreement was given by the so-called ‘peace dividend’15_{. Nonetheless, in a report prepared for the US Congress by}

Bolle (2001), two main elements were identified to justify the free trade agreement from an eco-nomic perspective. Firstly, the free trade agreement would sustain ecoeco-nomic growth and at im-proving the economic integration of the Middle East. Secondly, it would facilitate and promote further liberalisation reforms. As previously observed, this was eased by the fact that Jordan was

15_{The ‘peace dividend’ is a term used in political science to define the economic benefits originating from the}

ending of a war, largely determined by the reallocation of resources from the military sector to productive sec-tors of the economy (Ward & Davis, 1992).

(22)

18 a prominent example in the region for seeking to transform its economy into a market-oriented one.

The JUS FTA was prioritised after King Abdullah II bin Al-Hussein ascended King Hussein bin Talal, beginning its parliamentary procedure in February 1999. In spite of the strong and biparti-san support, labour and environmental provisions hampered the progression of the drafting. Opponents stated that additional provisions should not be included in the FTA, as the benefits from trade may not arise if the countries have to change their existing regulation (Bolle, 2001). Nevertheless, the JUS FTA garnered consensus and was signed on the 24 October 2000, enter-ing into force on the 17 December 2001. The agreement was the first of its kind with an Arab country, and it touched several aspects besides the elimination of tariffs and their degree of dis-persion, inter alia, protection of intellectual property rights, environmental legislation and issues

on labour. Figures 6 and 7 supply a first glance of the outstanding results of the FTA. On the left, figure 6 shows a decrease in the average tariff rate during the last decade, becoming virtually zero in 2015. On the right, figure 7 reveals a spike in Jordan’s export to the US starting in 2002, which increased by more than 20 times. As of 2014, the FTA appears to have been successful, as the US is one of the main trading partner of Jordan, with which the merchandise trade to the US amounts for 15.8% of its exports (WTO, 2015). However, the aim of diversifying its export may have not materialised yet. This can be easily seen by figure 8, which displays the breakdown of Jordan’s exports to the US. The stacked chart shows that the bulk of Jordanian exports is made up of textile products, whereas only chemicals and precious metals contribute to trade notably. Figure 6: Average tariff rate for imports from the US. The

graph shows the average for 68 types of weighted tariff rate. Values are indexed, so that 2000 = 100. Sources: TRAINS (UNCTAD) and author’s calculations.

Figure 7: Net Export and Net Imports from the United States. Exports and Imports are depicted by the continuous and dashed lines respectively. Values are gives as share of Jordan’s gross domestic product, which are computed from data in cur-rent US$. Source: World Bank.

0 20 40 60 80 100 2000 2003 2006 2009 2012 2015 Av er ag e Ta rif f R at e (2 00 0 = 10 0) 0.00000 0.00002 0.00004 0.00006 0.00008 0.00010 0.00012 1981 1992 2004 2015 Sha re of G D P (p er c ent )

(23)

19 Figure 8: Breakdown of Exports by Product. Values are given in US$ million. Colours of the writings of product types is related to the colour of the respective series. White writings are the only exceptions are they are related to the orange series that they overlap. Source: AJG Simoes, CA Hidalgo. The Economic Complexity Observatory: An Analytical Tool for Understanding the Dynamics of Economic Development. Workshops at the Twenty-Fifth AAAI Conference on Artificial Intelligence (2011).

Unfortunately, no microeconomic analysis has been undertaken on the effect of the Jordan-US FTA on poverty. 0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 Preparations of cereals, flour, fruits,vegetables Pharmaceutical products Inorganic and organic chemicals Articles of leather; saddlery and harness; travel goods, handbags Articles of apparel and clothing accessories, knitted or crocheted Articles of apparel and clothing accessories, not knitted or crocheted Other made up textile articles; sets; worn clothing and worn textile articles; rags Natural or cultured pearls, precious or semi-precious stones, precious metals Miscellaneous articles of base metal Aircraft, spacecraft, and parts thereof Miscellaneous manufactured articles

(24)

20

4. Methodology

Section 2.4 mentioned the obstacles that an empirical work faces when studying the relationship between free trade and poverty. This section presents how these obstacles are passed over, lay-ing out the methods employed to produce a poverty profile for Jordan and assesslay-ing how this one has been influenced by the FTA with the US. For the sake of exposition, I divide it into two main parts. Section 4.1. focuses upon the methodology used to construct the poverty profile, while section 4.2. specifies the empirical framework that aims at identifying the causal effect.

4.1. Construction of the Poverty Profile

The measurement of poverty is by itself an intriguing challenge, as it pursues to define in a quan-titative fashion a complex and multifaceted living condition. At the core of the topic there are three main problems:

• How do we measure household living standards? • What method do we adopt to define poverty? • How do we measure poverty across households?

Researchers have answered to these questions in several ways. However, since debate over the measuring of poverty is not the purpose of this work, I only present the most used method here-inafter.

The way to measure standards of living faces two problems, the first one being how to translate the wellbeing of individuals in a mathematical form that is consistent and comparable across people. This is not a trivial task, and the solutions found by researchers entail arbitrary choices whose validity is often harshly disputed. Such problem, at the base of the analysis, is determinant for the final outcome, as different measures of individual welfare might lead to drastically differ-ent results on the analysis of poverty (Ravallion, 2015). Without going into a detailed description of the school of thoughts of poverty measurement, I illustrate the main concepts of Welfarism and how their use in microeconomic theory applies to poverty analysis.

Assuming rational economic agents who are able to take part into the activities of the markets, individual welfare is identified by the bundle of goods and services people consume. This is the result of the maximisation process of one’s own wellbeing that involves a revelation of ordinal individual preferences through market prices. In this way, assuming that individuals derive direct utility from consumption, microeconomic theory proposes a measure of welfare that is grounded on this money-metric approach.

Current statistics make two measures eligible to be employed in this framework. The first is the real household consumption expenditure and the second is real household income. Scholars have generally declared the superiority of using real consumption expenditure over real income, as the former is measured more accurately than the latter and it has a closer relevance to poverty measurement (Ravallion, 1994). At the same time, real consumption expenditure is more closely associated to real permanent income (Glewwe, Gragnolati & Zaman, 2002), which provides a

(25)

21 more genuine parameter of household welfare. Therefore, this study follows the tradition of us-ing data on consumption expenditure.

The second issue regards what threshold of consumption expenditure one should set to define the poverty line, that is the upper limit above which a household is not considered poor. To de-termine Jordan’s poverty line, the DoS and the World Bank employed the Cost of Basic Needs (CBN) method (Ravallion, 2008). The CBN relies on the construction of a basket of goods which achieves an energy threshold of a specific calories per day16_{. While doing this, it is important to}

take into account local differences in price levels, that is to deflate prices across regions and are-as, but also to consider differences amongst the household members in terms of gender and age. This ensures the comparability between regional and national poverty lines.

The concept of poverty line implies that a household is poor if the following inequality holds true: 𝑐3,"∙ 𝑝3 6 378 ≤ 𝑐3∙ 𝑝3 6 378 (1) Where 𝑐 is the quantity of good 𝑛 consumed by household 𝑖, 𝑝 is its price, and 𝑐 is the upper-limit quantity defined by the poverty line. Thus, the element on the left hand side represents the aggregate basket of goods consumed by the household, while the element on the right hand side defines the aggregate consumption basket defining the poverty status.

Unfortunately, equation (1) cannot be computed, as precise quantities on food, clothes, and oth-er products consumed by the household are not available. This shortcoming is ovoth-ercome by em-ploying consumption expenditure instead, which implies the replacement of the quantity 𝑐 with its relative expenditure 𝑒. Another further modification regards the fact that, instead of using household expenditure, per capita expenditure is oft employed. This entails one additional steps, that is the introduction of the parameter ℎ𝑒 into (1), which represents the number of adult-equivalents in the household17_{. By considering 𝑒 has the total household consumption}

expendi-ture divided by the number of adult-equivalent members of the household, and 𝑒 as the respec-tive threshold of the poverty line, equation (1) becomes:

𝑒3,"∙ 𝑝3 6 378 ≤ 𝑒3∙ 𝑝3 6 378 (2) Having constructed an adequate poverty line, the last step regards the measurement of poverty, that is, to determine how far households lie from the poverty line. Typically, the literature

16_{The CBN is used by the DoS that, together with the World Bank, determine the amount of calories that are}

adequate in order to produce poverty estimates.

17_{As explained by Deaton (2003), the cost that households incur for children is smaller compared to that one of}

adults, and this has to be taken into account when computing per capita consumption expenditures. An ap-proach that is usually employed is to define the number of adult-equivalents as ℎ𝑒 = (ℎ + 𝛼 ∙ 𝑐), where ℎ is the number of adults, 𝑐 is the number of children, and 𝛼 is a parameter that represents the cost of a child relative to the cost of an adults. In the present study 𝛼 is equal to 0.5.

(26)

22 measures poverty by means of the three Foster–Greer–Thorbecke (FGT) indices (Ravallion, 2016), whose general formula is:

𝐹𝐺𝑇C= 1 𝑁 𝑧 − 𝑒" 𝑧 C 6 "78 𝑤𝑖𝑡ℎ 𝛼 = 0, 1, 2 (3) Where 𝑧 is the poverty line, 𝑒" is the real level of consumption of individual 𝑖, 𝑁 is the population

of the data, and 𝛼 is the parameter ranging from 0 to 2 that distinguishes the different indices. Equation (3) is a powerful tool, as by changing 𝛼 it is possible to concentrate upon different as-pects of poverty, like its severity and depth. If 𝛼 is equal to 0, the 𝐹𝐺𝑇J represents the poverty

headcount ratio. This measure is completely insensitive to the depth of poverty, so that it is not possible to appreciate the severity of households’ living conditions. Nevertheless, it is easy to understand as it gives at glance the share of population below the poverty line. If 𝛼 is equal to 1, the 𝐹𝐺𝑇8 is called poverty gap. This measure determines the average distance of the poor from

the poverty line, so that it provides useful insight regarding the distribution of poverty. Moreo-ver, it can be used as a toehold to calculate the cost for poverty alleviation. HoweMoreo-ver, it is not informative about the depth of poverty. Finally, when 𝛼 is equal to 2, the 𝐹𝐺𝑇K is called squared

poverty gap. This index presents the most desirable quantitative features for a poverty measure, as it permits to consistently compute the severity and depth of poverty. The 𝐹𝐺𝑇K is similar to

𝐹𝐺𝑇8, but differs in the weights given to poorer households, as the difference of those farther

away from the poverty line is emphasised. A thorough poverty profile requires the use of all three FGT indices.

4.2. Identification Strategy of the Effect on Poverty

4.2.1. Continuous Regression Model

Researchers have attempted to identify the effect of trade liberalisation on poverty either from a macroeconomic point of view or microeconomic one. Despite the usefulness of the concise re-sults provided by macroeconomic literature, it is likely that important phenomena arising at the microeconomic level are neglected. Hence, microeconomic analysis is more desirable (Winters, 2004).

Haughton and Khandker (2009) affirm that static regression analysis is the most effective tool in microeconomic analysis to establish the contribution of different variables to poverty. Two speci-fications belonging to this approach are explained by Glewwe, Gragnolati and Zaman (2002). On the one hand there is the continuous regression model, which analyses the impact of a set of ex-planatory variables on a continuous quantitative measure. In order to estimate the causal effect, this kind of models usually employ econometric techniques like difference-in-differences, or tools like instrumental variables (both topics are explained in section 4.2.2. and 2.4.3.). On the other hand there is the binary regression model, which estimates the effect of a similar set of ex-planatory variables on a binary qualitative measure, that is a dummy variable which determines whether the household is (or is not) poor. This approach employs multinomial logistic models.

(27)

23 Either model presents some desirable features and falls short other aspects. The method em-ploying a continuous dependant variable allows to identify more complex relations between consumption responses and explanatory variables, but it can struggle at categorising neatly be-tween poor and non-poor households. On the other side, the second method clearly defines between poor and non-poor households, but reducing a continuous variable to a qualitative var-iable may bring on a loss of information on the variation in the dependent varvar-iable with respect to the variation in explanatory variables. This issue is particularly worrisome if a larger number of households lies in nearby the poverty line. Also, Appleton (2002) demonstrates how discrete models are more suitable when dealing with nonlinearities than the continuous model.

Between the two methods, Haughton and Khandker (2009) state that the use of the continuous regression model yields better outcomes if one attempts to study the determinants of consump-tion expenditure. As described above, the binary approach does not allow to capture relevant information regarding the pattern of consumption, and this establishes the superiority of the continuous regression model for the present study.

The continuous regression model is employed in several studies, like the previously mentioned Hanson and Harrison (1999), McCaig (2011), Castilho et al. (2012), Justino et al. (2008), Le (2014), and many others. This methodology is also employed by Topalova (2007, 2010), whose frame-work is at the base of the present frame-work. In its most simple form, the continuous regression model develops from a simple linear specification that determines the real consumption expenditure of households from an explanatory variable:

𝑒",$= 𝛼 + 𝛽 𝑥",$+ 𝑢",$ (4)

Where 𝑒",$, the dependent variable, is the real per capita consumption expenditure of household

𝑖 in time 𝑡, 𝑥",$ is a time-varying explanatory variable, 𝛼 is a constant, and 𝑢",$ is an uncorrelated

error term with mean zero.

Equation (4) relies on a simple ordinary least squares approach, where 𝛽 gives the coefficient termining the effect of 𝑥 on 𝑒. As explained in section 2.4., the choice of 𝑥 is fundamental to de-termine the effect of the FTA on the dependent variable, and, since the reduction in tariff rates was certainly the main element of the agreement, the variable 𝑡𝑎𝑟𝑖𝑓𝑓 is the focus of this study. The Household Expenditure and Income Survey collected by the DoS allows to relate consump-tion expenditure to tariff rates thanks to a query that indicates the industry of employment of the household’s head. In this way, the equation (4) can be written as:

𝑒_",$= 𝛼 + 𝛽 𝑡𝑎𝑟𝑖𝑓𝑓_",$+ 𝑢_",$ ₍₅₎

4.2.2. Difference-in-Differences Approach

The implementation of the free trade agreement forced Jordan to dismantle its tariff barriers for goods imported from the United States. This implies that industries trading with the US were af-fected by the intervention and experienced a change in their profitability. Consequently, house-holds whose income stemmed from these industries possibly raised or lowered their

(28)

consump-24 tion expenditure according to the movements of wages, as seen in section 2.3. At the same time, the income of rest of the population that was not employed in the trading sector was arguably not affected by the FTA. This strategy is not uncommon in the literature and is applied, for in-stance, by Niimi et al. (2007).

Figure 9. Trends of the two groups without the free trade agreement. The subscript N indicates that the individual is in the non-treated group.

Figure 10. Trends of the two groups with the free trade agreement. The subscripts T and N indicates that the individual is treated non-treated respectively.

By taking advantage of this difference, equation (5) can represent the basic specification of a dif-ference-in-differences technique. The difdif-ference-in-differences is a technique characterised by the examination of two different groups in two subsequent periods, within the context of a natu-ral experiment. Individuals are either in the treatment group, which receives an intervention, or in the control group, whose contextual conditions do not change between both periods. By ob-serving the groups both at the beginning and at the end of the natural experiment, this model predicts that the treatment group will experience a larger (or lower) increase in the dependent variable with respect to the control group at the end of period 1. Figures 9 and 10 give a graph-ical representation of the phenomenon.

This approach implies that, instead of using the direct tariff rate faced by the sector of employ-ment of the household’s head, the explanatory variable is an employemploy-ment-weighted average of the nominal ad-valorem tariff rates enforced in the governorate where the household is living.18

By taking into account that household consumption expenditure is analysed with respect to this employment-weighted tariff, equation (5) becomes:

𝑒",$= 𝛼 + 𝛽 𝑡𝑎𝑟𝑖𝑓𝑓Q " ,$+ 𝑢",$ ₍₆₎

18_{The precise description and formula that is used to compute the employment-weighted tariff rate is given in}

section 4.2.3. E[e1,N,t0] E[e1,N,t1] E[e2,N,t0] E[e2,N,t1] t = 0 t = 1 e E[e1,N,t0] E[e1,N,t1] E[e2,T,t0] E[e2,T,t1] t = 0 t = 1 e 𝛽