The impact of agricultural policy distortions on the productivity gap: Evidence from rice production

Food Policy 36 (2011) 147–157

Contents lists available at ScienceDirect

Food Policy journal homepage: www.elsevier.com/locate/foodpol

The impact of agricultural policy distortions on the productivity gap: Evidence from rice production Manitra A. Rakotoarisoa ⇑ Trade and Markets Division, Economic and Social Development Department, Food and Agriculture Organization of the United Nations, Room D-835, Viale delle Terme di Caracalla, 00153 Rome, Italy

a r t i c l e

i n f o

Article history: Received 29 August 2008 Received in revised form 21 May 2010 Accepted 4 October 2010 Available online 3 November 2010 Keywords: Agricultural policies Trade distortions Rice productivity

a b s t r a c t This article explores how production and trade policy distortions affected rice productivity in 33 rice-producing countries. A rice-productivity index is constructed, and a model linking the productivity gap with policy distortions is presented. After controlling for the differences in infrastructure, access to inputs and equipment, openness, and human capital, this article shows that high levels of rice subsidies and protection in rich countries combined with taxation of rice farming in poor countries widened the gap in rice productivity between rich and poor rice countries. Ó 2010 Elsevier Ltd. All rights reserved.

Introduction The agricultural sector in many developing countries has suffered from both price distortions and low productivity, but attempts to find solutions have often overlooked the interaction between these two problems. For many commodities including rice, while developed countries have subsidized their production and exports and limited their imports, developing countries have taxed their producers. Although it is known that these distortions have depressed farm prices and revenues in small producing countries, their effects on farmers’ ability to adopt new technology and hence to increase productivity have received little attention.1 Because of depressed farm prices and low expected revenues, poor farmers may not afford the investments associated with the adoption of available technology, nor do they have the incentive to adopt technology to increase production. The link between agricultural policy distortions and productivity must be explored to help find ways to increase productivity and to reduce food insecurity, inefficient resource allocation, and lack of competitiveness in poor countries. The objective of this article is to determine how rice policy distortions affected rice-productivity gap for 33 prominent rice-producing countries during the period 1961–2002. The specific objectives are (i) to estimate the levels and growth rates of total

⇑ Tel.: +39 06 57053809. E-mail address: [email protected] Past studies (e.g., Nin et al., 2004) often assume that changes in agricultural policies are detached from productivity growth. One of the rare studies of the effects of government programs on productivity in agriculture is Makki et al. (1999), but it was confined to the US case. 1

0306-9192/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.foodpol.2010.10.004

factor productivity in rice production, (ii) to provide a model linking productivity with production and trade policies, and (iii) to estimate the impacts of policy distortions on productivity gaps among these producing countries. The innovation in this article is twofold. First, productivity indexes are constructed using a dynamic panel data model based on the model by Cermeño et al. (2003); this model specifies the residual measuring productivity in a two-way fixed effects autoregressive form and imposes neither stationarity nor constant returns to scale (CRS) technology on the production function. Second, an empirical model linking productivity to policy distortions is presented to explain why developed countries’ subsidies combined with developing countries’ taxes may have contributed to the productivity gap. This article suggests ways to rethink whether the usual recommendations to poor countries to improve their agriculture productivity and competitiveness are really detached from the need for policy reforms.

Rice policies and productivity The 33 prominent rice-producing countries, the focus of this study, are described in Appendix A (Table A1). Based on the differences in their agricultural value-added, their position as net exporters or importers, and their rice policies, these 33 countries can be roughly divided into three groups. Group 1 includes Organization for Economic Cooperation and Development (OECD) rice-producing countries: Australia, the EU (Italy), Japan, Korea, and the United States. This group has highest rice yields, heavily subsidizes its rice production and export, and sets high and various import protections. Group 2 includes developing countries that are net rice

148

M.A. Rakotoarisoa / Food Policy 36 (2011) 147–157

PSE Estimates for RICE % 100

80

60

%

40

20

0 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 year -20

-40

-60 India

OECD

Fig. 1. PSE estimates for rice (%). Sources: OECD (2004); Mullen et al. (2004), Thomas and Orden (2004); and Mullen et al. (2005).

exporters: Argentina, China, Colombia, Egypt, Guyana, India, Myanmar, Pakistan, Suriname, Thailand, Vietnam, and Uruguay. Countries in this second group mostly have relatively high yields and apply few producer subsidies but often tax their exports. Group 3 includes mostly low-income countries that produce rice but still import from the two other groups: Bangladesh, Brazil, Cambodia, Guinea, Indonesia, Iran, Laos, Madagascar, Mali, Malaysia, Nepal, Nigeria, Peru, Philippines, Sri Lanka, and Tanzania. Many countries in this third group have the lowest rice yields, and they often tax their rice production and export. Plots of the producer support estimates (PSE) in Fig. 1 show the stark contrast between the heavy production and export supports in developed countries in Group 1 and the low support and taxation of rice production in developing countries. The negative PSE values especially during 1991–1999 in India, for instance, indicate taxations on rice production. The developed countries’ producer support, an implicit export subsidy, and their various export programs have depressed prices received by farmers in small rice-producing countries (Wailes, 2004). Additionally, production and export taxes in these developing countries have depressed their farm prices further. Studies (e.g., Cramer et al., 1993, 1999; Wailes, 2004) agree that the removal of these distortions, especially tariff and subsidies in the highly distorted rice market, will lift international prices and improve overall welfare but with a skewed distributional impact. Studies, summarized in Tocarick (2008) warn of the skewed distributional impacts of the benefits of agricultural trade liberalization for low-income countries. For the rice sector, Wailes (2004) analyzes complete trade liberalization by rice categories and by group of countries and concludes that rice trade liberalization would lead to an increase in export prices (as a result of subsidy elimination), a decrease in import prices (as a result of tariff elimination) and an increase in aggregate welfare. But he also found that the distributional impacts are uneven across rice categories (long grain vs. short grain), countries (net exporters vs. net importers), and economic agents (producers vs. consumers). For long grain rice, producers in large importing countries, except those producing low quality grain rice, will gain. For medium/short grain rice (where price distortion is the most severe), the gain would be the highest

for producers in exporting countries (e.g., Egypt) although their consumers will lose. But for all rice categories, importing countries that had no tariff in place prior to the liberalization will lose because their consumers will be harmed by the rising price, although their producers may benefit from it. However, there have been claims (as summarized in Wise, 2008) that the farmers’ projected gains are overestimated, and that poor countries’ low level of productivity is one of the reasons preventing their farmers from benefiting from a freer agricultural trade. These arguments imply that the increase in farm productivity is often perceived as a detached prerequisite for benefiting from the reforms. But is it really the case? What if it were the distortions which depressed farm prices and revenues that discouraged poor farmers’ incentive to invest in technology (buy high-quality seeds, equipment) and to increase productivity and production? In other words, what if the removal of the distortions is what is needed to boost productivity? 2 This prompts the need to explore the link between the distortions and productivity to address these questions.

The price channel linking productivity and distortion As prices often capture and measure the extent of policy distortion in the market, they constitute also one of the channels that link policy distortions to technology adoption. To start with, the basic theory underlying such a view is that technology adoption generally increases with expected profits from the adoption of technology, i.e., it decreases with adoption costs and increases with expected farm revenue (Weiss, 1994). The expected profits depend mainly on prices received by farmers and prices of direct inputs associated with the technology. Indeed the work of Fulginiti and Perrin (1992) shows evidence that reinforces these views and concludes that taxes (or subsidies) on agriculture reduce (increase) 2 Authors remain divided. Some authors (e.g., Grossman and Helpman, 1991; Barro and Sala-i-Martin, 1995; Coe and Helpman (1995) Coe et al., 1997; Van Biesebroeck, 2003; Schor, 2004) argue freer and more open trade will lead to a productivity gain. But others in line with the Schumpeterian idea showed that protection helps increase saving and induces innovation, and as Rodrik (1992) shows, more imports do not necessary lead to a rapid technology catch-up.

149

M.A. Rakotoarisoa / Food Policy 36 (2011) 147–157

agriculture productivity: farmers expecting depressed output prices are reluctant to adopt new technology or to increase production. These ideas on the relation between output prices and innovation have been long established in the mainstream literature since the pioneering work of Schumpeter (1934) advocating that high market price often implies high profits and gives producers (or the sector) the resources and motivation to innovate and adopt new technology. The adoption of technology improves production efficiency and competitiveness, which helps expand profit further. Depressed output price, on the contrary, affects input demands by pushing the marginal value products of inputs below their competitive levels, hence, leads to a misallocation of resource and inefficiency. In agriculture, the same line of thoughts was echoed by Schultz (1956) and Smookler (1966) who were among the first to contend that agricultural output cannot be explained by an analysis based solely on conventional inputs and that policies that depress agricultural prices have a negative effect on agricultural productivity. Bingswanger (1978), Huffman and Evenson (1989) followed these leads and showed that agricultural prices serve as an incentive for innovation and for adoption of innovation. Schultz (1979a) also argued that the higher the price, the faster technology growth and in his Nobel Prize lecture, Schultz (1979b) emphasized: . . .The expectations of human agents in agriculture – farm laborers and farm entrepreneurs who both work and allocate resources – are shaped by new opportunities and by the incentives to which they respond. These incentives are explicit in the prices that farmers receive for their products and in the prices they pay for producer and consumer goods and services that they purchase. These incentives are greatly distorted in many low-income countries. The effect of these government induced distortions is to reduce the economic contribution that agriculture is capable of making. . . (Shultz, 1979b).

tion by the quantity of labor and taking the logarithms of both sides leads to the expression of the log of output per worker:

yit ¼

N1 X

aik xikt þ

k1

!

N X

aik  1 ln X iLt þ mit ;

k¼1

where the variables in lower case are the logarithms of the output or inputs per worker and XL is amount of labor. Based on the above expression the quantity yit  / yit1 can be written as yit  /yit1 ¼ ð1  /LÞyit ¼ ð1  /LÞ

N1 X

aik xikt þ ð1  /LÞ

k¼1

ð1  /LÞyit ¼ ð1  /LÞ

k¼1

N1 X

þli þ kt þ eit :

ð1Þ

Variables y and x represent the logarithms P of output  and input N per worker, and L is the lag operator; ai ¼ a  1 P 0. In (1) ik k¼1 the series is stationary if /  1 and the technology is CRS only if ai = 0. Two other alternative expressions of (1) are considered in replacing kt first by hi  t to allow the growth rate of productivity to vary across countries over time and then by h  t to indicate that productivity may grow over time at the same rate for all countries. Taking into account all these alternative expressions leads to the following econometric models of rice production per worker, for all i = 1, . . . , M:

yit ¼ /yit1 þ

N1 X

aik xikt þ

k¼1

N1 X

a0ik xikt1 þ ai ln X iLt þ a0i ln X iLt1

k¼1

þ hi  t þ li þ eit ; N1 X

aik xikt þ

k¼1

Framework and method

The rice production function for a country i (i = 1, . . . M) is specified in a Cobb–Douglas form and can be written in general terms Q a as Y it ¼ ð Nk¼1 X iktik Þ:emit , where the Y and X values represent output and inputs quantities. k (k = 1, . . ., N) represents an input type and t (t = 1, . . . , T) is time. The a’s are positive parameters and P k ak P 1; constant returns technology (CRS) is not imposed. The residual is exp(mit), where the residual mit represents the index of the level of technology to be estimated. Following Cermeño et al. (2003), henceforth CMT, I write the residual in a dynamic form as mit = li + kt + /mit1 + eit for which eit is iid (0, r2e ) where the l and k represent country- and time-specific effects, respectively. The terms l, k, and e are assumed to be uncorrelated to each other. / is the autocorrelation coefficient for the m series. The justification for the dynamic error component models stems from the heterogeneity of the cross-sectional units (countries) involved and the idea that over time, technology follows different paths for the various countries in the study. Dividing both sides of the production func-

aik xikt þ ð1  /LÞai  ln X iLt

k¼1

yit ¼ /yit1 þ

Productivity indexes

!

aik  1 lnX iLt þ ð1  /LÞmit

Using CMT expression of the residual and substituting ð1  /LÞmit ¼ li þ kt þ eit into the above equation leads to

This study borrows the ideas from these theories and evidences to explore the relation between rice productivity, especially riceproductivity gap among producers, and policy distortions.

I first provide a framework to construct indexes of productivity levels and growth. Then, I use a model that links policy distortions with productivity, employing these indexes to explain the productivity gap among rice producers.

N X

ð1aÞ N 1 X

a0ik xikt1 þ ai ln X iLt þ a0i ln X iLt1

k¼1

þ h  t þ li þ eit ; yit ¼ /yit1 þ

N1 X

aik xikt þ

k¼1

ð1bÞ N 1 X

a0ik xikt1 þ ai ln X iLt þ a0i ln X iLt1

k¼1

þ kt þ li þ eit :

ð1cÞ

where a0 ¼ /  a and a0 ¼ /  a, and the rest of variables and parameters are as defined above. The aim here is to determine which of the three specifications best represents the data to provide reliable estimates of the parameters h, l, and k; these estimates are employed to construct productivity level and growth indexes for each country (or group of country).3 Linking policy distortions to productivity gaps (a) Adoption of technology I borrowed the ideas detailed earlier that technology adoption generally increases with expected profits due to the adoption, i.e., it decreases with adoption costs and increases with expected farm revenue; the expected profits depend mainly on prices received by farmers and prices of inputs associated with the technology. It is assumed that the price received by farmers captures the distortions in the rice market. Besides, variables that may affect adoption of 3

This method differs from those in Ball et al. (2001) and Amadi and Thirtle (2004).

150

M.A. Rakotoarisoa / Food Policy 36 (2011) 147–157

technology also include variables linked to overall agricultural production such as infrastructure (e.g., dams for irrigation) and input delivery (Ransom et al., 2003), trade openness (reflecting the degree of trade-spillover effects on technology explained in Coe and Helpman (1995)), and human capital. Therefore, technology level reflected in level of productivity, TFP, can be expressed as: TFP = f(Pd, w, INFRA, HK, OPEN, FERT), or using the total difference expression:

dTFP ¼ dPd þ dw þ dINFRA þ dHK þ dOPEN þ dFERT

ð2Þ

where Pd is the price received by farmers, w represents the vector of input and equipment prices directly linked to the technology use, INFRA represents agricultural infrastructure index (e.g., irrigation and input delivery systems), HK is human capital, OPEN is degree of openness, and FERT is an index of essential input use (such as fertilizer). (b) Policies and productivity The next step is define Pd, the prices received by farmers. These prices differ for the three groups of rice-producing countries. Where Pri is the adjusted reference price (i.e., the adjusted world price at the location where competition takes place) for group i and Pdi is the domestic farm-gate price, the amount of money MPSi = Pdi  Pri is called market price support and measures the extent of the support programs. i. For Group 1 (OECD countries with high yields), the competition would occur at the national borders of the countries in Group 1 so that the following holds:

PR1 þ TR1 þ MM 1  CIF1; where PR1 is the adjusted reference price, TR1 is the cost of transportation from Group 1 members’ farms to their national borders, MM1 is the marketing margin between Group 1 farms and borders, and CIF1 is the rice c.i.f. price. The price received by farmers, which is relevant to revenue and profit for technological adoption, is Pd1 = Pr1 + MPS1, or by using the above identity,

Pd1 ¼ CIF 1  TR1  MM1 þ MPS1

ð3Þ

where CIF1, TR1 and MM1 are defined as above, and MPS1 is the price support. ii. For Group 2 (developing countries with relatively high yields), a net rice exporter, the competition lies at members’ borders (the exporting country’s borders) so that the following holds:

PR3  CIF 3 þ TR3 þ MM 3 ; where PR3 is the adjusted reference price, TR3 is the cost of transportation from farm to border, CIF3 is the rice c.i.f. price at Group 3 members’ borders (importing countries’ borders), and MM3 is the marketing margin between members’ borders and farms. The price received by farmers is Pd3 = PR3 + MPS3; the farm-gate price is then

Pd3 ¼ CIF 3 þ TR3 þ MM 3 þ MPS3

where preceding variable are defined as above, and MPS3 is the price support. Using (3)–(5), I can now define the differences between price received by farmers in OECD and farmers in the other two groups of countries:

dPd12 ¼ Pd1  Pd2 ¼ f ððCIF 1  FOB  TR1  MM 1  TR2 þ TR2 þ MM2 Þ; ðMPS1  MPS2 ÞÞ;

where PR2 is the adjusted reference price, is cost of transportations from Group 2 members’ farms to their national borders (the exporting country’s borders), MM2 is the marketing margin between Group 2 farms and national borders, FOB is the f.o.b. price for a competing exporter (an outsider), and TR2 is the cost of transportation from the competitor’s border to Group 2 borders. The price received by farmers is Pd2 = PR2 + MPS2, or by using the above identity

Pd2 ¼ FOB þ TR2  TR2  MM2 þ MPS2

ð4Þ

The preceding variable are defined as above, and MPS2 is the price support. iii. For Group 3 (developing countries with low yields), a net importer, the competition lies at the farm-gate so that the following holds:

ð6aÞ

dPd13 ¼ Pd1  Pd3 ¼ gððCIF 1  CIF 3  TR1  MM1  TR3  MM 3 Þ; ðMPS1  MPS3 ÞÞ

ð6bÞ

Eqs. (6a) and (6b) show that the difference in price received by rice farmers in OECD and in developing countries depends mostly upon (i) the wedges between the border prices of rice along with the differences between transportation costs and marketing margins on both sides of the trading groups and (ii) the contrast between the high price support in developed countries and taxation of rice production in developing countries.

Empirical model Because of symmetry and arbitrage, the first arguments of the function f() in (6a) and g() in (6b), namely the differences among cif and fob prices and among the marketing margins and transportation costs should vanish; if they do not vanish completely (because of the differences in price transmission mechanisms, market structure etc.) the remaining quantities can be captured as time- and country- (or group-) specific effects. The relevant term of the equations becomes then the second arguments, namely the differences in the levels of producer supports. Under such assumptions, the substitution of (6a), (6b) into (2) and the use of the log form of the difference lead to the following econometric model:

PR2 þ TR2 þ MM 2  FOB þ TR2 ; TR2

ð5Þ

ln

      TFP 1 OPEN1 HK 1 ¼ g0 þ g1 lnðS1  Si Þt þ g2 ln þ g3 ln TFPi t OPEN i t HK i t       INFRA1 FERT 1 w1k þ g4 ln þ g5 ln þ gk ln þ sit ; INFRAi t FERT i t wik t ð7Þ

where TFP is productivity level; S is a measure of domestic price support; OPEN, HK, INFRA, and FERT are the degree of country openness, the level of human capital, the agricultural infrastructure index (irrigation), and level of fertilizer use, respectively. The variable w is the price of agricultural and farm equipment. The coefficients gs are parameters, and s represents the error term. The subscript t refers to time; and the subscript 1 refers to Group 1 (OECD countries) and i to countries in either Group 2 (rice exporters in developing world) or Group 3 (rice importers in the developing world).

M.A. Rakotoarisoa / Food Policy 36 (2011) 147–157

Data and method

Results

Data are yearly and cover the period between the years 1960 and 2002. Appendix B explains the data and their sources. The estimation employed econometric methods and proceeded in two stages. The first stage was to estimate the three models 1a, 1b, and 1c. The variable inputs for the production function include labor, land, number of tractors, and fertilizer. The dependent variable is rice production per worker. I first pooled the data and considered the three groups of countries as the main cross-sectional units in order to compare the level and growth rate of productivity among the three groups. Then I estimated parameters within each group to obtain the technology-level indexes and growth rates over time. The second stage was to estimate the parameters of Eq. (7) using mainly the productivity indexes from the first stage and using agricultural support measure. OECD countries served as the group of reference, and all gaps were measured in terms of the ratio, (except for support S for which the gap is expressed in term of difference to avoid having a negative number inside the log), between average values for OECD countries and those of individual countries.

Measures of productivity level and growth

151

Tables 1–4 summarize the results of the estimation of rice-productivity indexes among and within the three groups of countries. For brevity, only the relevant parameters are reported here. Each of these tables shows the results of the estimation from three competing models based on the forms of the time component of the error terms to identify the level of technology; the model yielding the best fit, determined by the Akaike’s Information Criterion (AIC), is the bold-face column. Note that the coefficients on the lag of the log of output per worker are all significantly less than one, justifying the CMT specification and ensuring the stationarity of the series. (a) Between-group estimation In Table 1, the three groups of countries are taken as the cross-sectional units and Group 3 was chosen as the basis for comparison. In all three models, (1a)–(1c), the coefficients li for Group 1 and Group 2 are both positive and statistically

Table 1 Parameter estimates using the pooled panel data.

Note: The model (1a) has residual of the form li + hi  t; (1b) has li + h  t and (1c) has li + kt. Numbers in parenthesis and below the coefficients are t-values. Group 1 includes Australia, EU (Italy), Japan, Korea, and the United States. Group 2 includes Argentina, China, Colombia, Egypt, Guyana, India, Myanmar, Pakistan, Suriname, Thailand, Vietnam, and Uruguay. Group 3 includes Bangladesh, Brazil, Cambodia, Guinea, Indonesia, Iran, Laos, Madagascar, Mali, Malaysia, Nepal, Nigeria, Peru, Philippines, Sri Lanka, and Tanzania.

Table 2 Parameters of the rice production function for Group 1.

Note: The model (1a) has residual of the form li + hi  t; (1b) has li + h  t and (1c) has li + kt. Numbers in parenthesis and below the coefficients are t-values. Group 1 includes Australia, EU (Italy), Japan, Korea, and the United States.

152

M.A. Rakotoarisoa / Food Policy 36 (2011) 147–157

Table 3 Parameters of the rice production function for Group 2.

Note: The model (1a) has residual of the form l + h  t; (1b) has l + h  t and (1c) has l + k . i i i i t Numbers in parenthesis and below the coefficients are t-values. Group 2 includes Argentina, China, Colombia, Egypt, Guyana, India, Myanmar, Pakistan, Suriname, Thailand, Vietnam, and Uruguay.

Table 4 Parameters of the rice production function for Group 3.

Note: The model (1a) has residual of the form l + h  t; (1b) has l + h  t and (1c) has l + k . Numbers in parenthesis and below the coefficients are t-values. Group 3 includes i i i i t Bangladesh, Brazil, Cambodia, Guinea, Indonesia, Iran, Laos, Madagascar, Mali, Malaysia, Nepal, Nigeria, Peru, Philippines, Sri Lanka, and Tanzania.

significant and li is larger for Group 1. As expected, the high-income rice-producing countries in Group 1 have on average the highest level of productivity compared to rice exporters in Group 2 and to low-income rice-producing countries in Group 3. Based on model (1b), the coefficient h representing a common trend for all three groups indicates that the trend in world rice

productivity grew at 0.4% on average per year between 1961 and 2002. (b) Within-group estimation Tables 2–4 summarize the results within each group of countries. Table 2 presents the estimation on Group 1 (the OECD rice

153

M.A. Rakotoarisoa / Food Policy 36 (2011) 147–157 3.0000

2.5000

Index

2.0000

1.5000

1.0000

0.5000

01

97

99

20

19

95

91

93

19

19

19

89

19

87

19

19

83

85 19

81

19

79

77

75

Group 1 (high-income producers)

19

19

19

73

69

67

71

19

19

19

19

63

65

19

19

19

19

61

0.0000

Group 2 (low-income exporters)

Group 3 (low-income importers) Fig. 2. Rice-productivity index levels.

producers and exporters). Here one reason why (1b) produces the best fit could be that the countries in Group 1 are relatively homogenous in terms of the level of and increase in technology over the last four decades. Results show that Australia has the highest level of productivity index in this group: this is consistent with the country’s high level of paddy yields (see Appendix A). Moreover, the coefficient on trend, h, is significant and indicates that rice productivity in Group 1 grew at 1.6% per year on average for the last four decades. The coefficient on log labor is positive and statistically significant; rice in Group 1 is produced under slightly increasing returns to scale (IRS) technology; this is not surprising because in these high-income countries, technological advances and high level of supports often allow declining marginal costs. Table 3 summarizes the results in Group 2, net rice exporters among developing countries. Here, model (1a) is the best fit and shows that countries differ in both their levels of productivity and in their average annual productivity growth. The country-specific effects on productivity levels are all statistically significant in eight out of 11 countries. Countries with the highest levels of country-specific productivity indexes include Suriname, Uruguay, and Argentina. China and India have the lowest indexes in Group 2. Colombia, Uruguay, and Guyana have the fastest productivity growth. Compared with Vietnam’s, productivity growth rates in Pakistan and Thailand are relatively low. Here model (1b), though inferior to (1a), provides similar results. Moreover, results under model (1b) show that, overall, rice productivity for countries in Group 2 grew at almost 1% per year on average for the period 1961–2002. Result also shows that CRS technology assumption does not hold for Group 2. Table 4 presents the results for the low-income rice-producing and importing countries in Group 3. Here, as for Group 1, model (1b) best represents the data. All of the coefficients representing country-specific indexes of productivity levels are statistically significant; Peru, Indonesia, and Iran have the highest indexes. The coefficient on common time trend h is positive and significant, and indicates that, on average, rice-productivity growth for Group 3 was about 0.8% per year, almost half of that of Group 1. This result shows that the ‘catching-up’ hypothesis does not hold for rice production, as rice productivity for countries in Group 3 had started at low levels and grew at a slower rate than those of Group 1 and Group 2. The result on the return to scale is mixed, although under the best model, Group 3 produces rice under CRS technology.

(c) The productivity indexes and productivity gap Using the estimation results obtained so far, I constructed a time-series on estimates of the indexes of rice productivity for each country group. For this, I employed the group results under model (1a), which allows country- and time-specific effects for each time period. I used li as the starting productivity level and added kt level each year. The estimates of productivity levels are plotted in Fig. 2, which shows the widening productivity gap, especially between Group 3 and Group 1.4 The information on productivity indexes and gap among the three groups of countries will be used to estimate the link between policy distortions and productivity. The parameter estimates representing the productivity indexes are in general consistent with the reality of rice productivity in each group. To further check the validity of the estimated TFP indexes, I compared the annual growth rates of TFP indexes and actual yields and summarized the result in Table 5. Except for Group 1, the average growth rates of yields are higher than the average TFP growth rates. These differences are expected because yields are only a component of total factor productivity, and especially for Group 2 and 3, yields are average values over a large number of countries and their distributions present relatively high variations. Nevertheless the average growth rates of the yields and those of the productivity indexes are within reasonable and realistic range (all values are less than 2%). More importantly, the last row of Table 5 shows that the growth rates of yields and those of the productivity indexes are highly correlated. These results lend supports to the use of the productivity indexes for the rest of the analysis. Effects of policy distortions on productivity Table 6 presents the relevant parameter estimates of the model in (7) that links policy distortions and productivity. Complete data on PSEs were only available for the period 1982–2002 in eight rice-producing countries: Australia, China, India, Indonesia, Japan, Korea, Vietnam, and the United States. The coefficient of the producer-support gap is positive and significant at the 10% level for the entire period of 1982–2002 (second column, Table 6), indicating that the rice-productivity gap has increased as rich countries heavily subsidized and poor countries taxed (or only 4 The complete series of the productivity indexes for all the country groups between 1961 and 2002 are available but not reported here.

154

M.A. Rakotoarisoa / Food Policy 36 (2011) 147–157 Table 5 Rice yields and estimated productivity index: year-to-year average growth rates, 1960–2002.

Yields growth index per year (%) Productivity index growth per year (%) Correlation coeff. between yield growth and TFP index growth

Group 1

Group 2

Group 3

1.3 1.7 0.95

1.7 1.1 0.83

1.5 0.8 0.78

Note: Group 1 includes Australia, EU (Italy), Japan, Korea, and the United States. Group 2 includes Argentina, China, Colombia, Egypt, Guyana, India, Myanmar, Pakistan, Suriname, Thailand, Vietnam, and Uruguay. Group 3 includes Bangladesh, Brazil, Cambodia, Guinea, Indonesia, Iran, Laos, Madagascar, Mali, Malaysia, Nepal, Nigeria, Peru, Philippines, Sri Lanka, and Tanzania.

Table 6 Parameter estimates on the link between rice-productivity gap and producer supports. Independent variable: ln(TFP1/ TFPi) productivity gap

Coefficients

Dependent variables

No break

ln(S1  Si) Support price 1982–2002 Support price 1996–2002 Support price (1982–1995) ln(OPEN1/OPENi)a Openness ln(HK1/HKi) Human capital ln(INFRA1/INFRAi) Infrastructure-irrigation ln(FERT1/FERTi) Fertilizer use ln (w1/wi) Equipment price Observations: N  T = AIC

Uruguay-round break

0.015* (1.80) N.a. N.a. 0.102** (2.12) 0.021 (0.21) 0.617*** (11.09)

0.040*** (4.02) 0.015* (1.75) 0.104** (2.16) 0.019 (0.21) 0.619*** (11.17)

0.559*** (11.98) 0.587*** (9.95) 8  21 = 168  142

0.562*** (12.06) 0.588*** (-9.97) 8  21 = 168  76.1

Numbers in parenthesis and below the coefficients are t-values. a Openness is total trade over GDP. * Significance levels at 0.1. ** Significance levels at 0.05. *** Significance levels at 0.01.

slightly subsidized) their rice production and exports. This confirms the hypothesis that both heavy subsidies in rich countries and taxation of rice production in developing countries were associated with the rice-productivity gap. There are two possible ways that developed countries’ heavy support on rice particularly affected the rice-productivity gap. First, the level of support itself increased developed countries’ access to technology, hence to a strong productivity shift. Second, production subsidies, as an implicit export subsidy, depressed world prices and thus prices that farmers in small open developing countries received; low prices and revenues reduced these poor farmers’ incentives and abilities to invest in new technology. As a result the productivity gap between developed and developing countries widened. To check whether the estimate of the coefficient of the producersupport gap is sensitive to the implementation of the 1995 Uruguay Round Agreement on Agriculture, I divided the time period into two parts, before and after 1996. This split is also supported by data, which show that starting in 1996 the average PSE per ton in OECD countries started to fall, while developing countries continued to move from taxation to lightly subsidizing rice production.5 The results in Table 6, last column show that the coefficients of the producer-support gap remain positive, but their values and levels of significance are indeed different before and after 1996; the effect of the difference in the levels of support on the productivity gap is stronger and statistically more significant for the period after 1996. Before the start of the Uruguay Round Agreement on Agriculture, the increase of the gap in levels of producer support by 10% had increased the productivity ratio by 0.15%. This had necessarily

5

See OECD; USDA; Mullen et al. (2004), Mullen et al. (2005).

hindered developing countries’ efforts to access available technology during that period. After the Uruguay round, however, a 10% decrease in the difference between supports in developed and developing countries has lowered the index of productivity gap by 0.4%. This indicates that the reduction in the level of support in OECD countries and reduction of taxation of rice production in developing countries helped narrow the rice-productivity gap. In 2002, for instance, average rice PSE per metric ton of milled rice for Australia, US, Japan, and Korea was about 812 USD, whereas that of India was about 76 USD. For the same year, the calculated productivity index for OECD countries and for India were 2.388 and 0.968, respectively; the ratio (i.e., the productivity gap) was 2.467. The result implies that if the PSE gap had dropped by 50%, i.e., by 368 USD per ton (through any commitment to reduce producer support in the OECD countries), the productivity gap would have shrunk by (2.388/0.968)  0.04  50/100 = 0.0493 points. If the shrinking of the rice-productivity gap were entirely attributed to India’s increase in rice productivity, India’s index of rice-productivity level would have been 0.988, implying a 2% growth rate for its rice-productivity index.6 Such an increase is not negligible, taking into account how little the rice-productivity indexes have improved in the last 40 years for poor countries. For the impacts of the remaining variables, Table 6 shows that difference in the degree of Openness, proxied as the difference in the shares of total trade values in GDP, positively and significantly increases the productivity gap. This indicates that the spillover effects from trade on R&D and technology helped reduce the productivity gap, while trade protection hurt rice productivity in small open economies. This is consistent with the recommendations urging the poor countries to open up. The ratio of total trade to total GDP may, however, seem too broad a representation for openness in a sub-sector of agriculture like rice, and the results could be sensitive to the choice of proxy used to represent openness.7 For these reasons, I attempted to use as a proxy ‘rice openness,’ which is the ratio of the volume of rice traded to rice produced. Using this new proxy yielded a coefficient on openness of 0.013; although the coefficient remained significant at a level of 5%, it had become negative, meaning that increased protections in the rice sub-sector helped increase rice productivity. Although such a finding goes against the conventional idea that openness in poor countries through the R&D spillover effects should allow them to close the productivity gap, it is not entirely surprising for rice trade. The finding is consistent with arguments (e.g., Rodrik, 1992) that openness may not always guarantee closing of the productivity gap. More important, it is consistent with the reality in the rice trade because high-income countries, especially Japan and South Korea and to a lesser extent the United States, have achieved high growth rates of rice productivity, although the degrees of openness of their rice markets are among the lowest. One explanation is that for many developed countries, their high import protection has 6 India TFP index would have risen from 0.968 to 2.388/(2.467-0.493)=0.988, which is an increase of 2%. 7 See Edwards (1997) for explanations of the difficulty of choosing proxies for openness.

M.A. Rakotoarisoa / Food Policy 36 (2011) 147–157

kept their farmers’ income high, enabling these farmers to cover the cost of technology adoption. In contrast, small open countries’ rice farmers lost market share and income to the cheaper imported rice and could no longer afford the cost associated with technology adoption. In the end, the increases in both rice protection in developed countries and rice openness in developing countries led to the widening of the rice-productivity gap. Results in Table 6 also show that differences in access to agricultural equipment and in infrastructure variables linked to the overall agricultural sector (such as differences in the levels of irrigation and of fertilizer use) between developed and developing countries significantly affect the productivity gap. The only mixed result is on the coefficient of human capital variable; although the coefficient is positive as expected, it is not statistically significant. For the rest, a 10% decrease in the gap of the level of irrigation per unit of arable land is associated with a reduction in rice-productivity gap by 6%. Likewise, a 10% decrease of the gap in fertilizer use per unit of arable land would narrow the productivity gap by 5%. These results are not surprising. Many developing countries in Group 2 such as China and Indonesia have made significant efforts in improving agriculture infrastructure (such as irrigation dams) and input deliveries. These efforts have certainly increased productivity and eventually efficiency in agriculture and in rice farming. Similarly, results show that the difference in the access to agricultural equipment is one the sources of productivity gap.8

Conclusions and implications Using panel data on 33 rice-producing countries between 1961 and 2002, this article explores whether policy distortions have affected rice productivity. Productivity indexes were constructed using a dynamic panel model and were employed in the estimation of an empirical model linking the gap in producer support to the productivity gap between developed and developing countries. Results showed that levels of rice productivity have been significantly higher in high-income countries than in the rest of the world. The average rate of productivity growth in low-income rice-producing and -importing countries has been the lowest. There has been a sharp and widening productivity gap among the groups of rice-producing countries. Over the examined duration, the productivity growth rates were 1.7%, 1.1%, and 0.8% for the groups of rich countries, developing countries exporting rice, and developing countries importing rice, respectively; poor countries’ rice productivity did not catch up with the high level of rice productivity in rich countries. More important, results showed that the gap in the levels of support between rich and poor countries affected the rice-productivity gap. The high levels of protection and subsidy for rice farming in rich countries and taxation on rice producers in poor countries, especially before the Uruguay Round Agreement, widened the rice-productivity gap. But reduction of the support gap following the implementation of the agreement has helped narrow the productivity gap. Differences in agriculture infrastructure such as irrigation and in input delivery also affected the rice-productivity gap. Moreover, openness in total trade for developing countries helped reduce the rice-productivity gap, but high import barriers on rice in developed countries combined with openness in rice 8 I expect the sign on the equipment index to be negative because as the ratio ‘equipment price in developed country to equipment price in developing country’ increases, the productivity gap should decline. In other words, for farmers in developing countries, when the relative price of equipment falls (i.e. equipment becomes relatively cheap and more accessible), these farmers would be able to buy and use the equipment and close the productivity gap. This is consistent with the evidence in rice countries where reduction in tariff on imported equipment or imported inputs has boosted agricultural productivity.

155

trade in developing countries have widened the rice-productivity gap between developed and developing countries. The evidence would have been strengthened if additional data on rice production and rice policies were available for all countries. Also, the use of yearly average for many variables ignores the spatial and seasonal distribution of the values in each country; measuring the impacts of the policy distortions at the farm or community levels, for instance, is desirable. These shortcomings may have affected the findings, but the methods employed have opened avenues for future research on crop productivity. I draw three major implications from this study. First, reduction and removal of production and export subsidies and of high protection in developed countries will help rice producers from developing countries to increase productivity. This is because the removal of protection and subsidies entices higher prices for poor farmers, enabling them to earn enough to invest in new technology. Second, removal of developing countries’ taxations of rice production and export will increase incentives to adopt technology and increase productivity. Third, the rice-productivity gap will continue to widen if developing countries unilaterally open up their rice markets while the developed countries refuse to cut their protection and subsidies further. The findings and implications may extend to other highly distorted commodity markets such as maize, dairies, and sugar. The rice example highlights the need for comprehensive multilateral reforms that remove all distortions to help developing countries improve agriculture productivity. Increased productivity in developing countries need not be viewed as a prerequisite for benefits from the policy reforms but should be seen rather as a benefit emerging from such reforms. Acknowledgements The author thanks the editors and three anonymous reviewers for helpful comments and suggestions. The author is indebted to Jonathan Fowler for technical assistance and participants at World Rice Research Conference Tsukuba, Japan November 4-7, 2004 for their valuable comments. The views expressed in this paper are the author’s own and do not represent the official views of any organization. Appendix A See Table A1. Appendix B. Data B.1. Countries The 33 countries are divided into three groups depending on their levels of income and agricultural value-added per worker and their position as net exporters or net importers: Group 1: Australia, EU (Italy), Japan, Korea, and the United States. Group 2: Argentina, China, Colombia, Egypt, Guyana, India, Myanmar, Pakistan, Suriname, Thailand, Uruguay and Vietnam. Group 3: Bangladesh, Brazil, Cambodia, Guinea, Indonesia, Iran, Laos, Madagascar, Malaysia, Mali, Nepal, Nigeria, Peru, Philippines, Sri Lanka, and Tanzania. B.2. Variables The following refers to Eqs. (1), (1a)–(1c). Variable sources are in parenthesis

156

M.A. Rakotoarisoa / Food Policy 36 (2011) 147–157

Table A1 Rice production and trade in selected countries, 2002. Sources: United Nations Food and Agricultural Organization (FAO), 2005; US Department of Agriculture (USDA), 2005; World Bank, 2007a.

a

Country

Yields paddy rice (MT/ha)

Paddy production (MT)

Milled rice net export (MT)

Agricultural value-added per workera (USD 1995 constant)

1. Argentina 2. Australia 3. Bangladesh 4. Brazil 5. Cambodia 6. China 7. Colombia 8. Egypt 9. Guinea 10. Guyana 11. India 12. Indonesia 13. Iran 14. Italy 15. Japan 16. Korea 17. Laos 18. Madagascar 19. Malaysia 20. Mali 21. Myanmar 22. Nepal 23. Nigeria 24. Pakistan 25. Peru 26. Philippines 27. Sri Lanka 28. Suriname 29. Tanzania 30. Thailand 31. United States 32. Uruguay 33. Vietnam

5.746 8.607 3.423 3.324 1.916 6.186 5.011 9.141 1.613 3.825 2.683 4.469 4.727 6.270 6.582 6.350 3.086 2.141 3.090 1.971 3.674 2.675 1.024 3.018 6.687 3.280 3.489 3.940 1.964 2.609 7.373 5.863 4.590

713,449 1,291,000 37,851,000 10,457,100 13,822,509 176,342,195 2,346,940 5,600,000 845,000 443,700 107,600,096 51,489,696 2,888,000 1,371,100 11,111,000 6,687,225 2,416,500 2,603,965 2,091,000 710,446 22,780,000 4,132,600 3,192,000 6,717,750 2,118,616 13,270,653 2,859,480 163,410 640,189 26,057,000 9,568,996 939,489 34,447,200

220,745 268,540 942,872 531,998 17,227 1728,144 62,355 463,021 331,975 174,247 5,052,370 1967,717 868,789 504,619 626,902 151,299 26,400 61,082 518,205 10,076 723,744 14,636 1,199,637 1,670,491 34,175 1,196,157 92,784 42,975 67,475 7,336,663 285,701 650,876 3,201,000

10,374.550 36,865.740 322.218 5086.847 421.817 342.496 3636.244 1331.858 293.3051 4267.472 411.093 749.192 3790.597 27,654.190 34,140.070 14,743.210 624.339 156.128 6929.884 286.595 N.a. 206.351 742.223 698.227 1850.983 1475.778 710.0172 3619.626 190.289 878.126 53,402.960 7874.232 258.498

Year 2001 figures.

y = log of milled rice production in MT per 1000 workers (United Nations Food and Agricultural Organization (FAO), 2005); x’s = logarithms of inputs per 1000 workers: (i) Fertilizer: actual amount for developed countries; and the total use in times the share of harvested rice fields over arable land for developing countries; all in tons (FAO, 2005; International Rice Research Institute (IRRI), 2005). (ii) Physical capital: which is the number of tractor uses proxyied as the number of tractors per unit of land  areas harvested (FAO, 2005) (iii) Land: rice harvested area in hectares (FAO, 2005); XL = labor: nb. of workers in rice production proxyied as the rice-area share of agricultural labor force (FAO, 2005, World Bank, 2007a,b).9 The following variables refer to model in Eq. (7): TFP = TFP level index (Author); S = Producer support estimate based on computation of figures in USD/MT (OECD, 2004; United States Department of Agriculture (USDA), various years; Mullen et al., 2005; Nguyen and Grote, 2004; Thomas and Orden, 2004; World Trade Organization, 2007)10;

9 See Carter et al. (1999), Delgado and Chandrashekhar (1987). Also see Evenson et al. (1999) for an equivalent method. Basant and Fickert (1996) used total number of workers times the ratio of a firm labor cost to industry labor compensation to compute labor use in measuring an individual firms’ productivity. 10 See Tangermann (2006) for discussion about use of PSE as an index of agricultural support.

OPEN = i. Trade openness: Trade value over GDP; (Penn World Table by Heston et al. (2007), World Bank (2007a). ii. Rice openness: sum of import and export volume, divided by total production (FAO, 2005; USDA, various years; World Bank, 2007a,b); HK = Human Capital is the net secondary enrollment ratio (World Bank, 2007a); INFRA = Irrigation: ratio of irrigated land over arable land (FAO, 2005), = Fertilizer: fertilizer use in MT per ha (FAO, 2005, IRRI, 2005); w = Price index (deflated) of agricultural and farm equipment, or herbicide price index (FAO, 2005, IRRI, 2005; Ministry of Agriculture, Forestry and Fisheries of Japan (MAFF), Various years; US Department of Commerce, 2005; US Department of Labor, 2005).

References Amadi, J., Thirtle, C., 2004. Crop level productivity in the eastern counties of England, 1970–1995. Journal of Agricultural Economics 55, 367–383. Ball, E., Butault J., Nehring R., 2001. US agriculture 1960–1996, a multilateral comparison of total factor productivity. US Department of Agriculture, Economic Research Service, Washington, DC, Technical Bulletin No. 1895, May 2001. Barro, R., Sala-I-Martin, X., 1995. Economic Growth. McGraw-Hill, New York. Basant, R., Fickert, R., 1996. The effects of R&D, foreign technology purchase, and domestic international spillovers on productivity in Indian firms. Review of Economic and Statistics 78, 187–199. Bingswanger, H., 1978. Induced technical change: evolution of thought. In: Induced Innovation Technology, Institutions, Development. The Johns Hopkins University, Baltimore.

M.A. Rakotoarisoa / Food Policy 36 (2011) 147–157 Carter, C., Chen J., Chu, B., 1999. Agricultural productivity growth in China: Farm versus national measurement. Working Paper No. 99-001, Department of Agricultural and Resource Economics, University of California Davis. Cermeño, R., Maddala, G.S., Trueblood, M., 2003. Modeling technology as a dynamic error components process: the case of the inter-country agricultural production function. Econometric Reviews 22, 289–306. Cramer, G., Hansen, J., Wailes, E., 1999. Impact of rice tariffication on Japan and the world market. American Journal of Agricultural Economics 8, 1149–1156. Cramer, G., Wailes, E., Shui, S., 1993. Impacts of liberalizing trade in the world rice market. American Journal of Agricultural Economics 75, 219–226. Coe, D., Helpman, E., 1995. International R&D Spillover. European Economic Review 39, 859–887. Coe, D., Helpman, E., Hoffmaister, A., 1997. North–South R&D spillover. Economic Journal 107, 134–149. Delgado, C., Chandrashekhar, G., 1987. Technological change and agricultural labor use. In: Mellor, J., Delgado, C., Blackie, M. (Eds.), Accelerating Food Production in Sub-Saharan Africa. Johns Hopkins University Press, Baltimore, pp 118–134 (Chapter 9). Edwards, S., 1997. Openness, productivity and growth: what do we really know? Working Paper No. 5978, National Bureau of Economic Research, March. Evenson, R., Pray, C., Rosegrant, M., 1999. Agricultural research and productivity growth in India. Research Report No.109, International Food Policy Research Institute. Fulginiti, L.E., Perrin, R., 1992. Prices and Productivity in Agriculture. GATT Research paper 93-GATT2. Grossman, G., Helpman, E., 1991. Innovation and Growth in the Global Economy. MIT Press, Cambridge. Heston, A., Summers, R., Aten, B., 2007. Penn World Table Version 6.1, Center for International Comparisons at the University of Pennsylvania. Huffman, Evenson, 1989. The Development of US Agricultural Research and Education: An Economic Perspective. Staff Paper No. 174. Iowa State University Ames, December. International Rice Research Institute (IRRI), 2005. World Rice Statistic. . Makki, S., Tweeten, L., Thraen, C., 1999. Investing in research and education versus commodity programs: implications for agricultural productivity. Journal of Productivity Analysis 12, 77–94. Ministry of Agriculture, Forestry and Fisheries of Japan (MAFF), Various years. Statistical Yearbook of Agriculture, Forestry and Fisheries. Various issues. Mullen, K., Sun, D., Orden, D., Gulati, A., 2004. Producer support estimates for agriculture in developing countries: measurement issues and illustrations from India and China. Discussion Paper No. 74, Markets, Trade and Institutions Division International Food Policy Research Institute. Mullen, K., Orden, D., Gulati, A., 2005. Agricultural policies in India: producer support estimates 1985–2002. Discussion Paper No. 82, Markets, Trade and Institutions Division, International Food Policy Research Institute. Nguyen, H., Grote, U., 2004. Agricultural policy in Vietnam: producer support estimate: 1986–2002. Discussion Paper No. 79, Markets, Trade and Institutions Division, International Food Policy Research Institute. Nin, A., Hertel, T., Foster, K., Rae, A., 2004. Productivity growth, catching-up under uncertainty in China’s meat trade. Agricultural Economics 31, 1–16.

157

Organization for Economic Cooperation and Development (OECD), 2004. . Ransom, J., Paudyal, K., Adhikari, K., 2003. Adoption of improved maize varieties in the hills of Nepal. Agricultural Economics 29, 299–305. Rodrik, D., 1992. Closing the technology gap: does trade liberalization really help? In: Helleiner, G. (Ed.), Trade Policy, Industrialization and Development: New Perspectives. Clarendon Press, Oxford. Schor, A., 2004. Heterogeneous productivity response to tariff reduction: evidence from Brazilian manufacturing firms. Working Paper No. 10544, National Bureau of Economic Research. Schultz, 1956. Reflections on agricultural production, output and supply. Journal of Farm Economics 38, 613–631. Schultz, 1979a. Distortions of Agricultural Incentives. Indiana University Press, Purdue. Schultz, 1979b. The economics of being poor. Lecture to the memory of Alfred Nobel. Schumpeter, Joseph A., 1934. The Theory of Economic Development. Harvard University Press, Cambridge, MA. Smookler, J., 1966. Invention and Economic Growth. Harvard University Press, Cambridge. Tangermann, S., 2006. Response to the article on ‘How useful is the PSE in determining agricultural support? by Arie Oskam and Gerrit Meester. Food Policy 31, 142–147. Thomas, M., Orden, D., 2004. Agricultural policies in Indonesia: producer support estimates 1985–2003. Discussion Paper No. 78, Markets, Trade and Institutions Division. International Food Policy Research Institute. Tokarick, S., 2008. Dispelling some misconceptions about agricultural trade liberalization. Journal of Economic Perspectives 22, 199–216. United Nations Food and Agricultural Organization (FAO), 2005. FAOSTAT. . United States Department of Agriculture (USDA), Economic Research Service. Various years. Rice Outlook. United States Department of Commerce, Bureau of the Census. 2005. . United States Department of Labor, Bureau of Labor Statistics. 2005. Price index. . Van Biesebroeck, J., 2003. Exporting raises productivity in Sub-Saharan African manufacturing firms. Working Paper No. 10020, National Bureau of Economic Research. Wailes, E., 2004. Global trade and protection regime in rice trade. Working Paper, University of Arkansas. FAO Expert Meeting on Rice, November 2004. Weiss, A., 1994. The effects of expectations on technology adoption: some empirical evidence. Journal of Industrial Economics 42, 341–360. Wise, T., 2008. The limited promise of agricultural trade liberalization. Discussion Paper no. 19, Working Group on development and Environment in the Americas. World Bank, 2007a. World development indicators online. World Bank, 2007b. Commodity Price Data (The Pink Sheet). Development Policy Group, Washington, DC. World Trade Organization (WTO), 2007. Trade Policy Reviews: Secretariat and Government Various country reports. .