The Comprehensive Assessment of Water Management in Agriculture takesstock of the costs, benefits and impacts of the past 50 years of water developmentfor agriculture, the water management challenges communities are facing today,and solutions people have developed. The results of the Assessment will enablefarming communities, governments and donors to make better-quality investmentand management decisions to meet food and environmental security objectives inthe near future and over the next 25 years.

The Research Report Series captures results of collaborative research conductedunder the Assessment. It also includes reports contributed by individual scientistsand organizations that significantly advance knowledge on key Assessmentquestions. Each report undergoes a rigorous peer-review process. The researchpresented in the series feeds into the Assessment’s primary output—a “State ofthe World” report and set of options backed by hundreds of leading water anddevelopment professionals and water users.

Reports in this series may be copied freely and cited with dueacknowledgement. Electronic copies of reports can be downloaded from theAssessment website (www.iwmi.org/assessment).

If you are interested in submitting a report for inclusion in the series, pleasesee the submission guidelines available on the Assessment website or send awritten request to: Sepali Goonaratne, P.O. Box 2075, Colombo, Sri Lanka.

Comprehensive Assessment outputs contribute to the Dialogueon Water, Food and Environment Knowledge Base.

i

Comprehensive Assessment of Water Management inAgriculture Research Report 14

Multifunctional Agricultural Policy, ReducedDomestic Support and Liberalized Trade: AnEmpirical Assessment for Taiwanese Rice

Richard N. BoisvertandHung-Hao Chang

International Water Management InstituteP O Box 2075, Colombo, Sri Lanka

The authors: Richard N. Boisvert is, Professor, Department of Applied Economics andManagement, Cornell University, Ithaca, New York 14850, e-mail: rnb2@cornell.edu; andHung-Hao Chang is Assistant Professor, Department of Agricultural Economics, NationalTaiwan University, Taipei, Taiwan, e.mail: hunghaochang@ntu.edu.tw

Acknowledgments: The authors wish to thank the Council of Agriculture (COA), Taiwan,the Asian Rice Foundation and the Department of Applied Economics and Management,Cornell University for financial support and for the assistance of the Agricultural EngineeringResearch Center (AERC) in collecting data on rice production and the nonmarket valuesfor important non-commodity outputs.

The authors acknowledge also the assistance of Ranolph Barker, Gil Levine and DavidBlandford in the research and for comments on various sections of the report.

The authors accept responsibility for any errors or omissions. An earlier version of part ofthe empirical analysis was presented at the NAREA/CAES Workshop on InternationalTrade and the Environment, Halifax, Nova Scotia, June 23, 2004, and appears in Changet al. (2005).

Boisvert, R. N.; Chang, H. H. 2006. Multifunctional agricultural policy, reduced domesticsupport, and liberalized trade: An empirical assessment for Taiwanese rice. Colombo, SriLanka: International Water Management Institute. 44p. (Comprehensive Assessment ofWater Management in Agriculture Research Report 14)

Agricultural policy / international trade / rice / assessment / groundwater recharge / Taiwan

ISSN 1391-9407ISBN 92-9090-648-0ISBN 978-92-9090-648-3

Copyright © 2006, by International Water Management Institute. All rights reserved.

Cover photograph by Lucy Fisher, which is from a slide collection of the Department of PlantPathology Cornell University, Ithaca, NY, shows an area of terraced paddy fields in Asia.

Please send inquiries and comments to: comp.assessment@cgiar.org

The Comprehensive Assessment (www.iwmi.cgiar.org/assessment) is organized throughthe CGIAR’s Systemwide Initiative on Water Management (SWIM), which is convened by theInternational Water Management Institute. The Assessment is carried out with inputs fromover 100 national and international development and research organizations—includingCGIAR Centers and FAO. Financial support for the Assessment comes from a range ofdonors, including core support from the Governments of the Netherlands, Switzerland andthe World Bank in support of Systemwide Programs. Project-specific support comes from theGovernments of Austria, Japan, Sweden (through the Swedish Water House) and Taiwan;Challenge Program on Water and Food (CPWF); EU support to the ISIIMM Project; FAO; theOPEC Fund and the Rockefeller Foundation; Oxfam Novib and CGIAR Gender and DiversityProgram. Cosponsors of the Assessment are the: Consultative Group on InternationalAgricultural Research (CGIAR), Convention on Biological Diversity (CBD), Food andAgriculture Organization (FAO) and the Ramsar Convention.

iii

iii

Contents

Summary ........................................................................................................ v

Introduction .................................................................................................... 1

A Policy Framework for Multifunctional Agriculture and Other NTCs .......... 4

Conceptual Market Model Incorporating Multifunctional Policy Design ....... 9

Empirical Model ............................................................................................ 11

The Results of the Policy Analysis............................................................... 16

Policy Implications ........................................................................................ 21

Literature Cited............................................................................................. 23

Annex A ........................................................................................................ 27

Annex B ........................................................................................................ 33

v

To meet the requirements of membership in theWorld Trade Organization (WTO) as specified in2001, Taiwan must rely less on price supports andother traditional policies to support its agriculturalsector. In addition, it must open its agriculturalmarkets to increased imports. Central to thisreexamination of agricultural policy is the potentialconflict between domestic and international policygoals. For many, these so-called ‘non-trade’concerns (NTCs) are captured in the concept of‘multifunctionality’, which characterizes agricultureas a multi-output activity, generating bothagricultural commodities and non-commodityoutputs that are valued by society.

To address these new concerns, such as howfood is produced and environmental issues, anew paradigm is needed in which policyinstruments are oriented towards achieving theappropriate supply of both non-commodity andcommodity attributes of agriculture. Such aparadigm must be consistent with tradeiberalization and a distinction must be madebetween NTC’s that revolve around market failurein the production of multifunctional outputs, suchas landscape amenities and environmentalquality, and distributional issues, such as thelevel of farm income.

This report addresses the key elements of sucha new policy paradigm and its empirical significancethrough an examination of changes in Taiwan’s ricepolicy required for the country’s admission to theWTO. Through empirical simulations based on acomputable partial equilibrium model of theTaiwanese rice market, we examine the re-instrumentation of domestic policy to address bothpositive and negative environmental externalities,assess the implications of trade liberalization foroptimal policy choice and examine some issues in

farm income distribution. We model two of themultifunctional attributes that have figuredprominently in the debate on the environmentalimpact of rice production in Taiwan and other Asiancountries namely, groundwater recharge (a positiveexternality) and methane gas (a negativeexternality).

Through both our conceptual and empiricalanalyses, we have shown that a policy aimed atsecuring an appropriate supply of a major positiveattribute – groundwater recharge – and containinga negative attribute – methane emissions – wouldrequire that the price and income supportmeasures currently used be replaced by fees andcompensations for the use of inputs used in riceproduction. Recognizing that the specificnumerical results are sensitive to parametervalues, we have also demonstrated therobustness of the qualitative conclusions onappropriate policy design under substantialvariation in the valuation parameters forenvironmental attributes.

Net social welfare increases once theenvironmental externalities are internalized in thisway, and the gains are higher under tradeliberalization. This is consistent with the viewthat, overall, trade liberalization is welfare-enhancing. Furthermore, our results support earlyarguments by economists that trade policy shouldnot be used to correct domestic distortions. Inour context, the distortions created by the lack ofmarkets for the positive and negative externalitiesassociated with agricultural production are bestaddressed directly through domestic fees andcompensations for the use of inputs. And, at thesame time, these policies allow consumers tobenefit from lower product prices resulting fromreduced import protection.

Summary

vi

Furthermore, because the multifunctionalnon-commodity outputs are not produced in fixedproportion with output, our results also underscorethe fact that multifunctional objectives cannot bepursued efficiently through traditional price supportpolicy. The compensation of producers associatedwith the income support objective implicitly valuesgroundwater recharge below its social value, andunderestimates the social cost of methaneabatement.

There are direct policy implications for thedistribution of water between agricultural andindustrial uses in Taiwan. Achieving environmentalobjectives by replacing the current price supportpolicies would require the payment of land andwater subsidies and the imposition of a fee onfertilizer use. Under the valuations assumed forthe environmental goods and trade liberalization,water use would expand by about 8 percent sothat the value of groundwater recharge isincreased by about 7 percent. Fertilizer use woulddecline by about 8 percent. Land in rice productionwould rise 5 percent, partly to foster the increasein groundwater recharge.

Contrary to what one might expect, this policyre-instrumentation reduces rice production by justless than 4 percent. This is due in large part to thefact that the pre-WTO support price was paid onlyon a fixed quantity of production per hectare. Thus,even prior to policy reform, the supply-inducingprice at the margin was the domestic market-clearing price, rather than the support price.

Although there are gains in social welfare fromthe re-instrumentation of domestic rice policy inTaiwan, the efficiency gains are distributed towardconsumers and away from farm householdsproducing rice. Domestic payments to producersfall by 94 percent under trade liberalization. Theimputed land rents decline substantially, by anestimated 42 percent under trade liberalization.

Because of their magnitude, there areundoubtedly political, economic and equityreasons to address these distributional issues tofacilitate the process of adjustment to this newpolicy regime. However, these distributional issuesfall outside our definition of multifunctionality. Toaddress these distributional issues, somecountries have used compensatory payments to

producers to ease the adjustment processresulting from the reform of domestic agriculturalpolicies. Therefore, to gain some perspective onthe implications of providing adjustmentassistance, we assume that Taiwanese ricefarmers are compensated for nearly a 42 percentloss in land values due to the reduction in pricesupports and land set-aside payments. To do so,direct payments of about US$337/ ha. would berequired. The total cost, if this payment is madeonly to land in production, is US$117 million, or59 percent of the annual government outlays forprice supports and land set-aside payments. Ifdirect payments of this size were also made tothe remaining set-aside land, the cost would riseto US$158 million or 79 percent of the budgetoutlays under the traditional support policies.

In conclusion, although the results obtainedfrom this single empirical application may not be,generally, applicable to other countries andagricultural systems, the results of this empiricalinvestigation into the effects of the re-instrumentation of an agricultural policy in Taiwanare extremely encouraging. They demonstrate thatpolicy reform can contribute to optimal levels ofmultifunctional outputs, and it is also possible toaddress non-trade concerns that fall outside aworkable definition of multifunctionality. Directionsfor additional research are also suggested in theefforts made to reconcile existing social values orcosts of the multifunctional attributes of paddy.As is well understood in the environmentaleconomic policy arena, many of the multifunctionalattributes depend on local soil and water conditionsand other factors that affect agricultural production.The values placed on these attributes can differ aswell. Therefore, there is a need to extend thisresearch by considering explicitly these regionaldifferences. If they prove to be significant, thenpolicy re-instrumentation may have to beimplemented at the regional or local level. Inaddition, these types of policy analyses need to beextended to include additional multifunctionalattributes. To the extent that policies recognizethe contribution of farm production to a broaderrange of multifunctional attributes, the distributionalconsequences of policy reform may be diminished,thus reducing the need for adjustment assistance.

1

Multifunctional Agricultural Policy, ReducedDomestic Support and Liberalized Trade: AnEmpirical Assessment for Taiwanese Rice

Richard N. Boisvert and Hung-Hao Chang

Introduction

Rice is the primary food production industry inmany Asian countries, especially in Japan, Koreaand Taiwan. For years, the Taiwanesegovernment has paid close attention to the riceindustry and has employed severalmarket-intervention strategies to promote farmers’welfare and protect the food supply for the public.Rice policy evolved substantially during thesecond half of the twentieth-century in responseto changing domestic and international economiccircumstances. Between 1946 and 1973, thegoals of government rice policy were to stabilizeprices and expand agricultural supply. After thatperiod, the goals of government policy shifteddramatically—the primary focus was onincreasing farmers’ incomes through the adoptionof a price support program and other policies (seeannex A on page 24 for more details).

Since 1998, and as a condition for becominga member of the World Trade Organization(WTO), the Taiwanese government has had toreexamine the goals of its agricultural policy. Tomeet the free trade objectives of the WTO, thegovernment’s efforts to support its domesticagricultural sector must rely less on pricesupports and other domestic policies that distortinternational trade. Furthermore, along withdecreasing domestic support levels, thegovernment must open agricultural markets toincreased international trade.

Central to this reexamination of agriculturalpolicy (as Taiwan joins the WTO) is the potentialconflict between domestic and international policygoals. Since the conclusion of the UruguayRound negotiations under the General Agreementon Tariffs and Trade in 1994,1 the agriculturalpolicy debate has increasingly been dominated byso-called ‘non-trade concerns’ (NTCs). TheseNTCs reflect a broadening of agricultural policyobjectives beyond the traditional aim of providingprice and income support to farmers. A range ofconcerns has been identified, extending from therelationship between agriculture and theenvironment to the standards under which food isproduced, and the role of agriculture in ensuring asecure food supply (Blandford and Boisvert 2005).For many, these concerns are captured in theconcept of ‘multifunctionality’, which seeks toreflect the broader contribution of agriculture tosociety. Multifunctionality characterizes agricultureas a multi-output activity, generating bothagricultural commodities and non-commodityoutputs that are valued by society.

The reevaluation of agricultural policies fromthe perspective of their impact on the supply ofboth commodity and non-commodity outputsraises important questions about the extent towhich these NTCs can be addressed within thecurrent disciplines on international agriculturaltrade. For example, it has been common to view

1Signed in 1994, the Agreement on Agriculture (AoA) contains provisions acknowledging the critical linkage between domestic agriculturalpolicies and trade policies.

2

the supply of environmental attributes and othernon-commodity outputs as secondary factors inthe pursuit of traditional policy objectives, suchas income support. If addressed at all, the focuswould likely be on a single environmentalattribute. It has been argued elsewhere that anapproach that treats environmental aims and theproduction of other non-commodity outputs assubsidiary factors is an outdated policy paradigm(Blandford and Boisvert 2002). To address newconcerns, such as how food is produced andenvironmental issues, a new paradigm is neededin which policy instruments are oriented towardsachieving the appropriate supply of bothnon-commodity and commodity attributes ofagriculture. It is also important that such aparadigm be consistent with the liberalization ofinternational trade.2

Objectives

This report addresses the key elements of such anew policy paradigm and its implications. Theempirical significance of such reforms isinvestigated through an examination of changesin Taiwan’s rice policy required for the country’sadmission to the WTO. Through empiricalsimulations, we examine the re-instrumentation ofdomestic policy to achieve environmentalobjectives. We analyze policies that address bothpositive and negative environmental externalities,assess the implications of trade liberalization foroptimal policy choice, and examine some issuesin farm income distribution.

We focus on rice policy in Taiwan to illustratethe practical significance of these issues fordomestic policy reform and trade liberalization.Prior to joining the WTO in January 2002, Taiwanoperated an autarkic rice policy in which importswere prohibited. A price support program was inplace, combined with a land set aside to controlsupply. As a result of its membership in theWTO, a tariff rate quota (TRQ) was introduced to

permit limited imports of rice. Imports are madethrough a state trading enterprise (STE), whichcontrols their release onto the domestic market.The price support/set-aside program continues tobe used. Given that a degree of tradeliberalization has recently occurred, we are ableto explore the implications of a shift in domesticpolicy from price support to one in whichenvironmental objectives are paramount underconditions of autarky, and when the economy isopened to limited international competition.

Since a primary objective of the report is toidentify the policies that will lead to theproduction of the socially optimal levels ofmultifunctional outputs, one major difficulty inpolicy formulation is in determining appropriatesocial values for these nonmarket multifunctionalgoods. However, after a careful review of theempirical literature containing estimates of theseveral multifunctional attributes of paddy, itbecame apparent that there would be no way toreconsider the values for all multifunctionalattributes for inclusion in our empirical analysis.Therefore, we model two of the multifunctionalattributes that have figured prominently in thedebate on the environmental impact of riceproduction in Taiwan and other Asian countries,and the potential impact of reforms in domesticand international agricultural policies (Yang 2000;Lin et al. 2002; Tsai 1993; APO 2001). Thesetwo multifunctional outputs are: groundwaterrecharge (a positive externality) and methane gas(a negative externality) generated by paddyproduction. Although we attempt to reconcile theexisting wide range in estimates of the socialvalue of these two externalities in the literature,methods for valuing jointly produced nonmarketgoods are still under development. Therefore, inthe empirical analysis, we do investigate thesensitivity of the results to a range in theseenvironmental values.

In addition to ensuring that there are sociallyoptimal supplies of these multifunctionalattributes, the policy significance relates also to

2In contrast, and in an attempt to salvage traditional policy objectives, some have argued that the best way to sustain levels of thesenewly recognized jointly produced ‘non-commodity’ outputs is through a continuation of price supports. The research described in thisreport recognizes the critical need for an evaluation of these two alternative approaches.

3

the optimal allocation of land and water in theproduction of paddy relative to allocation of theseresources to competing nonagricultural uses.Finally, we address some issues related to farmincome distribution associated with thisre-instrumentation of Taiwanese agricultural policy.

Outline of the Report

Since a critical feature of this new policyparadigm is to design policy instruments orientedtowards achieving the appropriate supply ofmultifunctional, non-commodity attributes ofagriculture, we begin by developing a workingdefinition of multifunctionality. There can be littledoubt that the characterization of agriculture as amultifunctional industry has helped to broadeninternational debate; it has stimulated analysis ofthe issues and their implications for domestic andinternational agricultural policy. In bringing greaterclarity to the definition of multifunctionality, wemake transparent how multifunctional issues differfrom other important non-trade concerns (NTCs),and describe the implications for policy design inboth cases. In the empirical analysis, wedemonstrate how our policy solutions are not onlycorrect for market inefficiencies associated withthe supply of multifunctional non-commodityoutputs, but also address the distributionalconsequences of another important NTC—thedesire to support farm income or accommodatethe adjustment process associated with there-instrumentation of agricultural policy.

This discussion of multifunctionality isfollowed by the development of the theoreticaland conceptual models needed for the empirical

application. We first review some key conceptualissues relevant to policy design. These includethe relationship between agricultural productionand the supply of non-commodity attributes. Inthis regard, we pay particular attention to the factthat both commodity and non-commodity outputsare produced jointly, but not in fixed proportions.We use this review to develop an appropriateframework for the measurement and valuation ofjointly produced goods, which in turn serves as aguide for reconciling and refining the currentestimates of multifunctional values that appear inthe literature. This valuation framework, as wellas a discussion of important multifunctionaloutputs and other NTCs with a summary ofexisting value estimates, are described in greaterdetail in Boisvert et al. (2003) and Boisvert et al.(2004). We also discuss some conceptual issuesand data requirements for specifying jointproduction function relationships.

Next, to be able to isolate the effects of thisnew policy paradigm, we specify an empiricalmodel of the Taiwanese rice market, focusing onone positive externality—groundwater rechargeand one negative externality—methaneproduction, which are of particular policysignificance to both rural and urban residents. Inevaluating optimal policy choice, we giveparticular attention to instruments that areappropriate when it is difficult to observe andmonitor the supply of environmental attributesassociated with agricultural production.

The section that contains the empiricalspecification is followed by a discussion ofempirical results. The final section drawsimportant policy implications and outlinespriorities for continuing research.

4

A Policy Framework for Multifunctional Agricultureand Other NTCs

input that is fixed at the firm level; and(c) outputs are produced from a non-allocableinput (Boisvert 2001 a, b; Shumway et al. 1984;and Leathers 1991).

Case (a): Jointness due to technicalinterdependencies implies the presence ofeconomies or diseconomies in production. Anexample is joint production of honey and fruit,where fruit trees depend on insects for fertilizationbut also provide food for honeybees.

Case (b): In this case, the amount of a factorused in the production of each output can beidentified, but the total amount of that factoravailable to the enterprise is fixed in the shortterm. Another example that is illustrative ofmultifunctional agriculture would be the allocationof a fixed amount of land to several crops, aswell as to one or more open-space land uses thatprovide wildlife habitat or other environmentalbenefits (e.g., acreage enrolled in theconservation reserve program in the UnitedStates). The outputs from these land uses areinterconnected, because an increase in theamount of land devoted to one particular crop orto a particular conserving use reduces theamount of land available for the production of theother crops.

Case (c): This is the case where non-allocableinputs are used in the production of multipleoutputs (e.g., outputs are obtained from one andthe same input), and it may well be the mostrelevant of the three cases for the analysis ofmultifunctionality. The classical examples of jointproduction for agricultural commodities that fit thiscategory are meat and wool from sheep, and oiland meal from soybeans. Multifunctionalexamples would include such things as the jointproduction of milk and landscape amenities bygrazing cows on pasture and the joint productionof grain and nitrate leaching and runoff from the

From both technical and policy perspectives, it isimportant to distinguish between NTCs that relatedirectly to agricultural production (for example,environmental issues), and those that relate tothe changing position of agriculture in thedomestic economy (such as rural development orfood security issues).3 The policy problemsurrounding any particular NTC stems largelyfrom missing institutions or the failure of existinginstitutions. Failure to distinguish among differenttypes of NTCs, or the use of an unduly inclusivedefinition of multifunctionality, blurs the logic ofpolicy intervention and hinders the search forappropriate policy solutions. The real danger isthe appearance that a ‘one size fits all’ policyapproach is appropriate, when in reality, this islikely to result in a ‘one size fits none’ outcome.

A Workable Definition ofMultifunctionality

To group the NTCs that result from similarinstitutional failures and for which there arecommon policy remedies, a workable definition ofmultifunctionality must be restricted to thosenon-commodity outputs of agriculture that satisfytwo conditions. The first is that the non-commodity outputs must be jointly produced withthe food, fiber and materials that are thecommodity outputs of agriculture. The second isthat these non-commodity outputs must providesocial values (or impose social costs) that arenot reflected in organized markets.

Joint Production

In a formal sense, jointness in production canarise in three distinct ways: (a) there aretechnical interdependencies in the productionprocess; (b) outputs compete for an (allocable)

3This section, as well as others in the report, draws heavily on work by the senior author in collaboration with David Blandford. Inseveral recent papers, they have articulated the need for a new agricultural policy paradigm that recognizes the social value ofmultifunctional outputs from agriculture. (e.g., Boisvert 2001a, b; Blandford and Boisvert 2002; Blandford and Boisvert 2005).

5

use of commercial fertilizer on cropland.Because we cannot separate the contributions ofcows and pasture to milk production andamenities, production is joint. It is alsoimpossible to disentangle the contribution offertilizer to:a) the production of grain; and, b) tonitrate leaching or runoff.

To illustrate the complexity of multifunctionalagriculture, if we count the pollution from animalwaste, there is a third joint product in ourlivestock farming examples. Similarly, landscapeamenities resulting from the use of croplandwould add a third joint product to the grainfarming example in that the land’s contribution tograin production cannot be disentangled from itscontribution to amenities.

Moreover, these examples also underscoreanother important feature of the joint,multifunctional production of commodity andnon-commodity outputs—they are unlikely to beproduced in fixed proportions with the agriculturalcommodities (OECD 2001). As is seen below,this characteristic of multifunctional agriculturehas important implications for policy.

The Specific Case of Taiwanese Rice: In thecase of Taiwanese paddy, importantmultifunctional attributes of rice production arealso the result of non-allocable inputs used in theproduction of multiple outputs. In the currentpolicy debate, a number of importantmultifunctional attributes of paddy have beenidentified. The most frequently mentioned relateto: (a) the recharge of underground aquifers;(b) the amelioration of land subsidence; (c) thereduction in flooding; the reduction in soil erosion;(d) change in water and air quality; and(e) landscape amenities.

As suggested by the detailed discussion ofthese multifunctional attributes in Boisvert et al.(2003 and 2004), the major emphasis is on thosemultifunctional attributes that have positive socialvalue rather than those that impose costs onsociety. However, it is important to recognize bothkinds of multifunctional attributes and, ideally onewould include the entire list of multifunctionalattributes in any agricultural policy analysis. Todate we know of no attempt at such a

comprehensive policy analysis. As is alsoapparent from the detailed discussion on therange of multifunctional attributes, such ananalysis is also impossible here, partiallybecause of a lack of information on agriculturaloutputs and inputs to the levels of thesemultifunctional outputs. Also there are inherentdifficulties in assigning social values and costs tojointly produced nonmarket goods.

It is primarily for this reason that in ourcurrent examination of Taiwanese rice, we focuson one important positive multifunctional output—groundwater recharge, and one negativemultifunctional output— the negative effects onair quality from green house gas emissionsassociated with the release of methane gas frompaddy production. It is easy to see that thesenon-commodity outputs are indeed jointlyproduced with rice. Since groundwater recharge isdirectly related to total land planted to paddy riceand the intensity of the application of irrigationwater, it is impossible to disentangle thecontribution of these two inputs to the productionof rice from their contribution to groundwaterrecharge. Similarly, the contributions of land,water and fertilizer to rice production cannot beseparated from their effects on the release ofmethane. For example, methane released duringrice production depends importantly on the watersupply system; deepwater rice fields generatesignificant amounts of methane. Methaneproduction also depends on soil quality, soiltemperature, fertilizer practices and rice variety(e.g., Lin et al. 2002; Yang 2000).

Social Value or Cost

The second condition for a non-commodity outputto fall within our definition of multifunctionality isthat it must provide value to society (or impose asocial cost) not reflected in markets. There aretwo reasons why such nonmarket outputs haveno prices.

First, the joint non-commodity output may bewhat economists label ‘public goods’—anindividual’s enjoyment (consumption) of such agood does not reduce the quantity available to

6

others, and it is not possible to exclude someonefrom consuming the good once it is madeavailable (Baumol and Oates 1988). Since theconsumption of these goods is nonexclusive,consumers have no incentive to reveal the valueplaced on them. This public-good (nonexclusive)nature of multifunctional agriculture perhaps bestcharacterizes landscape amenities, culturalheritage and the effects of CO2 emissions onglobal warming.

Second, the joint non-commodity output couldbe in the form of a technical externality—definedin economic terms as a cost or benefit that is aby-product of economic activity (agriculturalproduction in this case) allocated outside themarket system (Baumol and Oates 1988). Unlikepublic goods, positive and negative externalitiescan be exclusive and depletable, but they arealso un-priced. The environmental effectsassociated with agricultural production (such asnitrate leaching and runoff from the jointproduction examples above) are perhaps bestcharacterized as negative technical externalities;they impose external costs on society.Alternatively, by raising the water table orincreasing the outflow from aquifers to springs,streams, or the sea and reducing the threat ofland subsidence, groundwater recharge frompaddy production generates external benefits tosociety that are not captured in the market.

For both public goods and technicalexternalities, farmers have no incentive to takeaccount of the external costs or benefits tosociety of jointly produced nonmarket goods inmaking production decisions. The consequence inboth cases is that, from society’s perspective,resources are misallocated. For example, byhaving no incentive to account for the ‘positive’social value of landscape amenities, farmers mayallocate too little land to farming activities forwhich such amenities are high. On the other sideof the ledger, if farmers do not consider

environmental implications of production decisions(for example, the contamination of water byagro-chemicals or animal waste or the release ofmethane to the atmosphere), crop or livestockproduction in certain areas may exceed thesocially optimal levels.

The Optimal Policy Solution

By defining multifunctionality in this way, theoptimal policy solution is to internalize the effectsof the externalities or ‘public’ good attributes ofthe non-commodity outputs, i.e., to ensure thatthey are taken into account in farmers’ resourceallocation decisions. The standard economicprescription is to ‘compensate’ producers by anamount equal to the marginal social value of apositive externality and ‘charge a fee’ equal to themarginal social cost of a negative externality. Inso doing, we obtain the social welfare maximizinglevels of non-commodity outputs (Baumol andOates 1988).4

In an earlier paper on the subject, Plott(1966) emphasized the importance of taxing orsubsidizing the correct thing, as illustrated below.His observations are particularly significant inlight of the belief held by some that appropriatelevels of multifunctional outputs can be achievedthrough a continuation of traditional domesticprice supports. Although not dealing withagriculture per se, Plott recognized explicitly thatthe environmental externality was produced jointlywith another conventional output. He was alsowriting at a time when explicitly or implicitlythrough particular examples, many believed thatthe corrective fee or compensation should beplaced on the production of the commodity.However, Plott demonstrated that such a policy(e.g., a price support on the commodity in thecase of a positive externality) would beappropriate only under extremely restrictive

4While these are the well-known Pigouvian subsidies and taxes, Blandford and Boisvert (2005) argue that it is important to avoidreferring to appropriate compensation and fees in terms that may have a pejorative connotation. Since the non-commodity outputsare of value to society, compensation is, in a very real sense, a payment to producers for services provided, rather than a means ofredistributing income. In the case of a negative technical externality, producers are being charged a fee in lieu of the cost theirproduction activities impose on society, rather than simply to generate government revenue. Throughout the report, the terms feesand compensations are used interchangeably with the terms taxes and subsidies when discussing Pigouvian policy instruments.

7

conditions i.e., there be only one input used inproduction and that the externality be produced infixed proportions with the commodity output.5

He went on to show that the fee or compensationshould be assessed on the externality itself, orunder certain conditions on the inputs used in theproduction of the externality. This latterobservation proves critical in our empiricalapplication.

Multifunctionality as a Basis for Policy

To understand why this definition ofmultifunctionality and the policy prescriptionsabove provide a sound basis for furtherdiscussions of agricultural policy formation, wemust know what sets these multifunctionalnon-commodity outputs apart from other non-tradeconcerns such as food security, and ruraldevelopment.

These other non-trade concerns relateprimarily to whether freer trade undermines theability of a country to guarantee a sufficientsupply of safe food or to promote the economicand social development of rural areas. They arealso qualitatively different from technicalexternality issues associated withmultifunctionality. The changes in response topolicy reform that give rise to these NTCs do notinvolve missing markets. Instead these concernsare reflected through the effects of the levels ofactivity and prices in other markets (e.g., rurallabor markets, agricultural input markets, etc.).

Accordingly, these non-trade concerns fallunder the heading of pecuniary externalities—theeffects of production or transactions on outsideparties through resulting changes in othermarkets and prices (Baumol and Oates 1988;Kolstad 2000). Although other market outcomesare affected, pecuniary externalities do not leadto a misallocation of resources, but there maybe redistributive effects associated withtransfers among different segments of society.

These types of non-trade concerns may also beaffected by the impact of freer trade ondomestic agricultural production. While notrequiring any public action because there is noresource misallocation, there may be politicalincentives to address the distributionalconsequences. Such concerns should not beaddressed through the price system, but ratherthrough institutions put in place to ensureappropriate distributional outcomes (for example,domestic decoupled payments to maintain farmhousehold incomes and a viable rural economy;government stocks to provide security in theavailability of food; food standards andinspection systems to protect food safety). Tounderscore the differences in institutionalsolutions for these NTCs and multifunctionality,policies to deal with the distribution of farmincome under policy re-instrumentation are alsoaddressed in the empirical analysis below.

Challenges to ImplementingMultifunctional Policies

Although the standard economic prescriptions tomultifunctional policy, to ‘compensate’ producersby an amount equal to the marginal social valueof a positive externality and ‘charge a fee’ equalto the marginal social cost of a negativeexternality, appear simple enough on thesurface, there are three major challenges to dealwith when implementing. The first relates tovaluation of jointly produced non-commodityoutputs. The second relates to difficulties inobserving or monitoring the levels ofmultifunctional outputs, while the third relates toestablishing an unambiguous relationshipbetween commodity production, agricultural inputuse and the production of non-commodityoutputs. We discuss each of these issuesbelow, and their practical solutions with respectto Taiwanese rice are discussed as part of theempirical specification.

5 If these very restrictive conditions are not met, a tax or compensation on output could well make matters worse, rather than better.

8

Valuing Nonmarket Goods

Assigning monetary measures to the benefits orcosts associated with nonmarket goods is alwaysa challenge and, environmental economists haveused both direct and indirect methods to do so.

There are three basic indirect methods thatrely on observed choices: (i) the hedonic priceapproach exploits the notion that the price ofgoods, such as a house or a job, can bedecomposed into the prices of the variousattributes of the good (e.g., the air qualitysurrounding a house in addition to the size of thehouse, the number of rooms, etc.);(ii) alternatively, the weak complementarityapproach makes use of the fact that theconsumption of some goods can becomplementary to changes in environmentalquality. For example, if an improvement inenvironmental quality results in an outward shiftin the demand for visits to a lake, then the valueof the improvement can be measured by the netaddition to consumer surplus under the newdemand curve; and (iii) the averting behaviorapproach is perhaps the most relevant for ourpurposes. This approach is based on the fact thatin some cases expenditures on other inputs orgoods and services can be used to mitigate theeffects of pollution or other externality. If this isthe case, then the value of a 'small change' inpollution can be measured by the value of theinputs (or other goods and services) used tocompensate for the change in pollution. Variationson this approach might also be called theavoidance cost method (e.g., the water filtrationcosts avoided by a household because of aspecific improvement in water quality), or thereplacement cost method (e.g., placing a valueon additional water catchment by calculating thecost of additional reservoir capacity to serve thesame purpose). Using the costs of abatement, asis done for estimating the social cost of methaneemissions in our empirical application, is yetanother variation on this theme.

The alternative to these indirect methods isto gather direct evidence on the value ofnonmarket goods through surveys based on thecontingent valuation method (CVM) and discrete

choice modeling based on conjoint surveyresponses, or through experiments conducted in alaboratory for experimental economics. Randall,one of the pioneers of the CVM methods,summarizes the intuition behind these methods inthe following way. Paraphrasing, if we designquestions with enough care, perhaps people canprovide reliable evidence of amenity values bytelling us their (willingness to pay) WTP or(willingness to accept) WTA directly; or by tellingus what they would do (e.g., buy or not buy)given well-specified choice situations that weconstruct for them (Randall 2002).

Measurement Issues

Regardless of whether a direct or indirectapproach is taken, the valuation process isparticularly difficult in the case of agriculture’ssupply of jointly produced multifunctionaloutputs. The challenges begin with what toinclude in the list of nonmarket goods and howto measure or articulate the characteristics to bevalued. These challenges are easily illustratedusing some of the multifunctional benefits ofpaddy listed above.

The measurement issues are really difficultwith respect to items such as landscape. Forexample, although the public may be willing toplace a subjective value on landscape, whatexactly are the characteristics of landscape thatare valued? Is it a particular pattern of land use,small fields for example? Is it that there areparticular crops in the fields or that the fields areinterspersed with wooded areas that providewildlife habitat?

There are also some measurement issuesrelated to groundwater recharge and methaneemissions, etc., but for these multifunctionaloutputs, the critical factors are perhaps morerelated to identifying the physical relationshipsbetween agricultural inputs and the ‘joint’production of the commodity and non-commodityoutputs. A second concern relates to difficultiesin observing or monitoring the levels ofnon-commodity outputs.

Some of these relationships are currentlybased on sound engineering and knowledge of

9

biological processes, but empirical application isoften difficult because of a lack of data needed toestimate or evaluate these relationships. Theseissues are particularly difficult to resolve inempirical analysis if the relationships depend ondifferences in soil, climate and other regional-specific factors.

By knowing these physical relationships, weare also able to circumvent much of the difficultyassociated with the non-observability of thenon-commodity externalities. We do so bytargeting multifunctional policy intervention atobservable variables that are correlated with theexternality generating process such as the inputsused for agricultural production (e.g., Rude 2001;Romstad 2004). In the absence of marketdistortions, the first-best welfare scenario can stillbe achieved if the appropriate fees(compensations) are applied to ‘all' inputscontributing to the production of thenon-commodity externalities or ‘public’ goods(Plott 1966; Holtermann 1976; Griffin and Bromley1982).

Valuing Jointly Produced Nonmarket Goods

Regardless of whether the multifunctional policiestarget the multifunctional outputs directly or the

inputs essential to their production, our optimalpolicy design relies on individual values for eachof the multifunctional outputs. And, regardless ofwhether these values are estimated using director indirect methods, problems can arise in usingseparate values when the several non-commodityoutputs are 'jointly produced'. Hoehn andRandall (1989) and Carson et al. (1998) cautionthat problems can arise in this case if the netsocial benefits are derived by adding together theindividual values of the separate nonmarketgoods (e.g., non-commodity outputs in our case),each being derived independently and individuallyusing conventional valuation procedures. Byimplicitly ignoring the effects of ‘joint production’and policy interactions, such procedures willsystematically overstate the benefits orunderstate the costs. Some proposed strategiesfor addressing this issue within the context ofmultifunctional agriculture are described byRandall (2002), but to date, no empirical workhas been attempted. To facilitate the empiricalanalysis at hand, this issue is also addressedspecifically by Boisvert et al. (2003) in modifyingthe use of existing values of groundwaterrecharge based on contingent valuationestimates of several multifunctional outputs foruse in this study.

Conceptual Market Model Incorporating MultifunctionalPolicy Design

To conduct the empirical analysis ofmultifunctional policy for Taiwanese rice, anumber of these challenges mentioned above hadto be addressed to the extent possible, giventime and other resource constraints and somelimitations on the availability of data. Theresolution of each issue is described as part ofthe empirical specification.

The optimal policy formulation built into thisempirical model is based on a conceptual modelincorporating multifunctional policy design. Thecomplete formulation of this conceptual model isin annex B. Before moving on to a description of

the empirical model of the ‘Taiwanese RiceMarket,’ it is helpful to summarize the keyfeatures of the conceptual model and highlight therelationships between the social values and costsof the multifunctional outputs, and the levels ofthe policy instruments.

An essential component of this conceptualmodel needed to derive the optimalmultifunctional policy design is the specificationof a general transformation function for both riceproduction and the two non-commodityenvironmental outputs (equation 1B of annex B).As mentioned above, it is most appropriate in this

10

case to think of the joint production being due to anon-allocable input. Therefore, one cannotdisentangle the separate contribution of each inputto each product, and the total input use is notdetermined by summing inputs used by eachproduct as it would result in double counting.

Both characteristics of multifunctionalagriculture affect modeling farm-level and marketbehavior. If markets are competitive and if weignore the social value or cost of non-commodityoutputs, a farmer would maximize profits byequating the marginal value product of each inputwith its price. If it is possible to observe thelevels of production of the non-commodityoutputs, then we also know that to maximizesocial welfare, the standard economic prescriptionis to ‘compensate’ producers by an amount equalto the marginal social value of a positiveexternality and ‘charge a fee’ equal to themarginal social cost of a negative externality.

However, it is difficult or impossible tomonitor the levels of the two multifunctionaloutputs in this study (e.g., groundwater rechargeand methane emissions). Therefore, following thetheoretical results mentioned above, we derivesocially optimal policies with appropriate fees(compensations) being applied to 'all' inputscontributing to the production of thenon-commodity externalities.

In this report, we extend a similar conceptualmodel by Peterson et al. (2002) to the casewhere there are also market distortions due totraditional agricultural policies such as pricesupports and land set-asides. The appropriatemodifications in the levels of the input fees(compensations), for what is now a second-bestoptimum due to the domestic or internationalpolicy intervention, are derived in a mannersimilar to that when the market distortion is in theform of imperfect competition.

Two-stage Approach to Policy Design

As discussed in annex B, it is convenient to designoptimal policies either in two steps or in stages. Inthe first stage, the government implements theoptimal policy design by taxing or subsidizing land

and other inputs. Given the levels of the policyinstruments announced by the government, therepresentative rice-farmer would make optimaldecisions on the use of agricultural inputs in asecond stage. Since a solution is derived throughbackward induction, it is convenient to begin with adiscussion of the second stage.

Optimal Decisions of Producers

The representative rice-farmer is assumed tomaximize profits (revenue minus costs, equation2B in annex B). Revenue includes market salesof rice, and, under the traditional domesticsupport policies, rice sold to the government atthe support price and payments for the set-asideland. Costs include the sum of the inputquantities used multiplied by the input marketprices. For purposes of accounting for the socialcosts or benefits of the multifunctional outputs,the farmer’s profit equation also includes thecompensations or fees for the use of inputsassociated with the production of themultifunctional outputs.

It is well known in economics that tomaximize profits, the farmer would use inputs upto the point where the additional revenue from theuse of an additional unit of input is equal to theprice of the input. This is no longer true, however,under an optimal multifunctional policy design.Under these conditions, the farmer equates themarginal revenue from the use of an additionalinput with the input price, adjusted for themarginal ‘net’ social benefit due to the use ofanother unit of the input. If the ‘net’ socialbenefit is positive (negative), input use under thepolicy will be higher (lower) than it would be underthe private optimum, and the levels of themultifunctional outputs will also be higher(lower) than the levels associated with the privateoptimum.

Optimal Levels of Environmental PolicyInstruments

The optimal levels of the compensations and feesfor the use of inputs are, in turn, derived bymaximizing a social welfare function, representedby the sum of consumers’ and producers’

11

surpluses, less governmental budget costs, plusthe total benefit associated with the positiveenvironmental externality and minus the totaldamage from the negative externality (equation5B of annex B).

These optimal compensations and fees arederived from the optimality conditions formaximizing this social welfare function. Theoptimal input compensation (fee) is determined inpart by the marginal contributions of each input tothe production of the externalities multiplied bythe marginal benefit or damage of eachexternality. If more than one input contributes toboth externalities, it is impossible to determine'ex ante' if the compensation (fee) is positive(negative). For example, the optimal fee for anon-land input may well be negative (for exampleimplying compensation) if that input’s marginalcontribution to the benefits associated with thepositive externality outweighs its marginalcontribution to the cost of damage associatedwith the negative externality. Whether thecompensation for land is positive or negativedepends on similar considerations in our model,but, if the traditional policies are also still inplace, the size of the land compensationdepends in part on the distorting effect of thelimited price support and the land set-asidepayment. To underscore the effect of marketdistortions on the level of ‘Pigouvian-type'policy instruments, the land compensation thatensures the optimal level of the multipleexternalities will decrease if the level ofdomestic support (either the price support or the

set-aside payment) increases (equation 10Bbfrom annex B). Without domestic support, theoptimal compensation for land is equal to the neteffect of land’s net marginal contribution to bothexternalities, just as it is for other inputs.

Opening the Economy to InternationalTrade

The optimal environmental policy designrepresented in equations (10Ba and 10Bb) fromannex B applies to a closed economy, and wasapplicable to the rice market prior to Taiwan’sadmission to the WTO, which, in 2003, led to riceimports under a tariff rate quota (TRQ). For thisreason, we must also determine how the optimalmultifunctional taxes and subsidies are affectedby the new policy regime.

To understand the TRQ, for a smallimporting country, (e.g., the situation for rice inTaiwan), we discuss three possible outcomes inannex B. Under the first outcome, where importsequal the minimum access commitment, theTRQ acts like a quota in which a tariff is alsolevied. As is evident in the empirical analysisbelow, it is this case that is applicable toTaiwanese rice. Accordingly, the optimalenvironmental policy for this small openeconomy is similar to that from above, exceptthat the domestic price is now determined by thesum of domestic production plus imports, ratherthan by just domestic production as in theclosed economy case.

Empirical Model

To illustrate the economic impact of policychanges, a computable partial equilibrium modelis used to represent the Taiwanese rice market.This framework has been widely used foranalyzing the effects of agricultural policies(e.g., Floyd 1965; Maier 1991; Gardner 1987).A special feature of the approach is that thevarious market levels in the vertical

production/consumption chain are consideredsimultaneously. The approach is adopted primarilybecause agriculture represents only about2 percent of Taiwan’s domestic product—thus itwould be unlikely that there would be noticeablegeneral equilibrium effects from changes in ricepolicy. Furthermore, similar partial equilibriummodels have been used elsewhere to examine

12

the effects of TRQs and the reduction indomestic agricultural support (e.g., Abbott andPaarlberg 1998; Boughner and de Gorter 1999).Finally, as demonstrated above, the optimalenvironmental policies can be analyzed inconjunction with the adoption of a TRQ bymaking some rather straightforward changes inthe partial equilibrium framework. In order toisolate the impact of trade liberalization, weexamine scenarios for both a closed economyand an open economy. We benchmark ourempirical model assuming a closed economy withthe domestic policies in place as in 2001. Theessential features of the Taiwanese rice marketare captured in 13 equations (table 1).

Because data to estimate the parameters ofinput supply equations and other suchrelationships are not readily available, we employ

reasonable estimates of the parameters derivedfrom the literature, relying particularly on studiesfrom Japan, other Asian countries and the UnitedStates of America. Since a primary focus of thereport is on optimal environmental policies, wefocus our sensitivity analysis of the empiricalresults on the range in estimates ofenvironmental values.

Input Supply Equations

Equations (1) through (4) represent input supplies:farmland (L), farm labor (Z), fertilizer (FP) andirrigation water (W), respectively. Research forthe Japanese rice industry (Ohba 2001) is thebasis for the value used for the supply elasticityof land, 0.55. The value is towards the low end

α

α

αα

αααα

(1) L=gl(pl )=kl Pl

(2) Z=gz(pz )=kz Pz

(3) FP=gfp(Pfp )=KfpPfp

(4) W=gw (pw) =kwPw

(5) PFl+ Q(P-P)+(1- ) Ps+ sl =Pl

(6) PFz = Pz

(7) PFfp = Pfp + t

(8) PFw + sw = Pw

(9) F ( L,Z,FP, W) = k ( L)C1 Z

Cz FPCfp W

Cw=Qd

(10) PP = kp P

(11) PP = a - b (Q d + M)

(12) GW = kgwWTPlog {W dw ( L)

d1}

(13) ME = kmrW ewFP

e fp ( L)e1

β β

εl

εz

εfp

εw

TABLE 1.

The equations for the model of the Taiwanese rice market.

13

α

of the range of 0.0 to 2.0 found in previousliterature (Floyd 1965; Gardner 1987). Sinceindividual irrigation associations have controlover the allocation of irrigation, and the transferof land in and out of agriculture has to beapproved by a government sponsored, county-level farmers’ organization, we believe it isappropriate to assume the supply of land isrelatively inelastic.

In past studies for the United States ofAmerica, estimates of the elasticity of laborsupply to agriculture have ranged from 1.0 to3.0 (Gisser 1971; Rosine and Helmberger 1974).A study by Tyrchniewicz and Schuh (1969)found labor supply elasticities ranging fromaround 0.7 in the short term to around 1.5 in thelong term. We assumed a value towards the lowend of this range, 0.8, for the Taiwanese riceindustry.6

Supplies of purchased inputs, such asfertilizer, are usually much more elastic thanthose of other agricultural inputs. The range offertilizer supply elasticities found in the literatureis from 0.5 to 10 (Gardner 1987). We assumed asupply elasticity of 2.0 for the Taiwanese riceindustry.

We found no studies that provide empiricalestimates of the supply elasticity of water forirrigation, but irrigation associations control mostof the irrigation water used in Taiwan’s riceproduction. Therefore, it seemed appropriate toassume that the supply of irrigation water wouldbe quite unresponsive to the imputed value ofwater for rice production—an elasticity of 0.3 isused in our analysis.

Input Demand Equations

Equations (5) through (8) are the derivedinput demand equations for farmland, labor,

fertilizer and water, respectively. Each variable isas defined in the theoretical model, and theadditional term, Fi, is the partial derivative withrespect to input i of the Cobb-Douglas productionfunction, which is embedded in the market clearingequation (9). Three of the inputs (fertilizer, landand water) contribute to the production of the twoenvironmental externalities. Demands for thesefactors are affected by fees or compensationsassigned to them by the government (t, sl and sw,respectively). The demand for land is also affectedby the level of price support, the set-asidepayment and the proportion of land that is setaside. Based on data published by the Council ofAgriculture (CoA 2002), the proportion of landsuitable for rice that is actually allocated to riceproduction ( ) is around 0.7. The government’spurchase price for rice ( p ) was US$621/ton; themaximum amount that the government wouldpurchase at this price was 1.26 ton/ha. The set-aside payment per hectare (Ps) was US$1,213/ha.(CoA 2002). These policy parameters andenvironmental fees and compensations are set atvarious levels, or eliminated altogether in the policyscenarios described below. In calibrating the model,the domestic policy variables are set at their 2001levels and the environmental fees andcompensations are set to zero.

The Production Function, Rice Demandand the Market Equilibrium

Equation (9) is the market clearing condition forrice production (the terms on both sides of the firstequals sign) and consumption. We were somewhatlimited in our choice of rice production functionsbecause only time-series data on the cost of riceproduction were available. This is the primaryreason for specifying that the production functionfor rice has a Cobb-Douglas form. However, the

6 We recognize that these estimates of the supply elasticity of labor are dated, and we did search the literature for more recentestimates—to no avail. We did, however, perform some sensitivity analysis and determined that estimated changes in net socialwelfare and all other important variables from the model are extremely insensitive to changes in the supply elasticity for labor. Thereare two explanations for this result. The first, and perhaps the most important, is that all policy impacts are compared to a 2001baseline. To be consistent with this baseline, the model is initially calibrated using the assumed values for factor supply elasticities. Allpolicy impacts are measured relative to this calibrated baseline. The second reason for the relative insensitivity of the results is that,as is seen below, labor is the only factor that does not contribute to the production of the environmental goods, and thus it is neithertaxed nor subsidized in the optimal environmental policy.

_

14

In the monsoon environment, rice is grown insaturated soil that is often flooded throughoutmost of the growing season. In addition toproviding optimum growing conditions, theproduction system affects the naturalenvironment. It may affect the frequency andintensity of flooding, groundwater recharge, soilerosion and the water and air quality. In addition,there may be important social and economicexternalities relating to landscape and recreation.

There was certainly no way to model allmultifunctional attributes of rice. To be able todemonstrate the interaction between environmentalpolicy design within the context of domestic andinternational trade policy reform in agriculture, itwas necessary to focus on a single importantpositive externality (groundwater recharge) and asingle important negative environmental externality(methane emission) associated with riceproduction. These externalities are chosen not onlybecause of their particular interest in Taiwan’sagricultural research and policy circles, but alsobecause it was possible to reconcile existingestimates of their nonmarket values.

The Positive Externality: GroundwaterRecharge

One major consequence of the ponded conditionsof paddy production is the percolation of water intothe soil. The rate of percolation varies, dependingupon soil type, but water moving downward has arecharge effect on groundwater. This will replacewater withdrawn from the aquifer, raise the watertable or increase the outflow from the aquifer tosprings, streams, or the sea (Ohnishi andNakanishi 2001). It may also reduce substantiallythe risk of land subsidence (Barends et al. 1995).Recharge is influenced strongly by the status ofthe underlying aquifer. If the aquifer is in overdraft,i.e., the phreatic surface is declining over time, therecharge water may, according to some, beconsidered to have a value equal to that of waterstored above the ground. If the aquifer is notstressed by the rate of extraction, recharge maysimply sustain the base flow of springs andstreams fed from the aquifer. This may have

Cobb-Douglas form has performed well in otherstudies where production functions wereestimated from Taiwanese rice data (Tsai andWann 1995). We used time series data on thecost of rice production between 1952 and 2001 toestimate the parameters of the productionfunction. Average cost shares over that period forfarmland, labor, fertilizer and water were: 0.19 (cl),0.57 (cz), 0.22 (cfp), 0.02 (cw).

In the market closing condition, we make thecommon assumption for the relationship betweenvertical market levels in the industrial organizationliterature (Tirole 1988) that there is a fixedproportional yield of table rice from raw rice.Based on the conversion rate published by theCoA (2002), the value of ��in equation (9) is setat 0.7. Equation (10) defines a fixed relationshipbetween the farm price and the consumer pricefor rice.

Equation (11) is the linear demand functionfor rice. The consumer price is determined by theamount of rice available in the domestic market—domestic production in the autarky case and thesum of domestic production and imported rice (M)released onto the domestic market in the with-trade case. Parameter b is calibrated based onthe assumption that the consumer demandelasticity for rice is -0.1 (Yang and Chen 2000).The low demand elasticity reflects the fact thatrice is the main staple food in Taiwan; given therelatively high level of consumer income, demandis insensitive to changes in the price of rice.

Modeling the Multifunctional Attributes

Although non-trade concerns have traditionallyfigured in the debate on agricultural trade policy,the range of non-trade concerns has broadeneddramatically in recent years. Much of this hasbeen associated with the characterization ofagriculture as a multifunctional activity. In thisregard, Taiwan’s identification of non-commodityoutputs in rice production is no exception. Someof the issues identified in Taiwan are common tothe production system used to grow ricethroughout monsoon Asia.

15

)()( 3/2L

L

W

αα

significant environmental value, e.g., maintainingfish populations, but the value is likely to be lowerthan that for an aquifer in overdraft.

To model groundwater recharge, it wasnecessary to specify how recharge relates toinput use. Based on an agricultural engineeringstudy, groundwater recharge is assumed to bedirectly related to total land planted to rice andthe intensity of the application of irrigation water(Matsuno et al. 2002). We model groundwaterrecharge as a semi-logarithmic function, equation(12); an increase in the intensity of theapplication of irrigation water per unit of land areawill affect groundwater recharge, as will anincrease in overall land use. We can see this bywriting the term in braces { } from equation (12)

as { }, which is also equal

to . Groundwater rechargeincreases with the use of both inputs (land andwater), but at a decreasing rate. The value ofgroundwater recharge is reflected by multiplyingthe right-hand side of the equation (12) by anestimate of the nonmarket value, WTP.

The estimates for the value of groundwater(WTP) are derived from values reported by Chen(2001) and Chen et al. (2002). In both studies, CVMmethods are used to estimate the value ofgroundwater recharge associated with Taiwaneserice production. However, while Chen et al. (2002)focus only on the groundwater recharge value, Chen(2001) estimates the value of groundwater jointlywith the value of other multifunctional outputs ofrice production. Thus, it is difficult to disentanglethe effects of groundwater recharge in the latterstudy. To reconcile the two estimates of value, thetotal values were converted to marginal values.These estimates were weighted averages, theweights determined by the proportion of thepopulation in households of different types and indifferent counties (Boisvert et al. 2003; and includedreferences). The average willingness to pay for aone percent change in groundwater recharge wasestimated to be US$844, with an upper bound ofUS$1,114 and a lower bound of US$574. The rangeof estimates of WTP for groundwater recharge is

the basis for a sensitivity analysis.

The Negative Environmental Externality:Methane Emissions

Although nitrate residuals are among themost widely recognized sources of agriculturalpollution, numerous test results (e.g., from datasupplied by the Kaoshung Irrigation Association)suggest that current levels of nitrates inTaiwan’s drinking water are within acceptablelevels of concentration. For this reason, wefocus on another issue—methane emissionsfrom rice fields.

Methane is produced during rice productionby aerobic decomposition of soil organic materialin flooded rice fields. Almost 90 percent ofmethane generated and oxidized by aerobicbacteria in the soil reaches the atmosphere,thus contributing to greenhouse gas emissions.In 1990, it was estimated that rice contributed16 percent of total methane emissions worldwide(US EPA 1999). One important factor affectingmethane generation from rice production is thewater supply system. Deepwater rice fieldsgenerate significant amounts of methane. Someof the methane bubbles up through the water, butmost reaches the atmosphere by traveling up therice stalk through the plant’s vascular system(Hyman 2001). Other factors that affect methaneemissions from rice production are soil quality,soil temperature, fertilizer practices and ricevariety. For example, if farmers use biogasresidues instead of barnyard manure in riceproduction, methane emissions can be reducedby 24 percent to 62 percent. If farmers use hybridrice seed, methane emissions can be mitigatedby 10 percent (ADB 1998).

Based on the nature of this aerobic process,we assume that fertilizer, water and land used inrice production contribute to methane emissions,as in equation (13). This specific formulation isbased on estimates of total methane emissionsby Yang (2000) and Lin et al. (2001). Averagetotal emissions from rice production in these twostudies were estimated to be 35,500 tons.Further, from a recent study of methane

3/13/2 )(αLW

16

abatement in China (ADB 1998), it is estimatedthat a one percent change in rice production willchange methane emissions by roughly 2 percent.By combining these sources of information withthe input production elasticities from theproduction function, we can relate methaneemissions to input use. The details are inBoisvert et al. (2004).

The negative value of methane emissions tosociety is estimated as the product of averageabatement cost (r in equation [13]) and totalemissions. Our estimates of average abatementcosts depend on the available technology for the

The Results of the Policy Analysis

abatement and total decrease in the amount ofmethane. According to a recent study of riceproduction in China, the most efficient abatementstrategy is through effective manuremanagement, involving an abatement cost ofUS$86/ton; a much less efficient strategy isthrough the adoption of hybrid rice, with anabatement cost of US$1,342/ton. Anotherpossible strategy is through the use of a drynursery for which the abatement cost is US$402/ton (ADB 1998). Again, these values are used inevaluating the sensitivity of our results withrespect to the social cost of methane emissions.

columns B and E show what happens if thegovernment keeps its existing price and incomesupport instruments, but also usescompensations and fees to address theenvironmental externalities.

The key conclusions that can be drawn fromthese results are:

1. Achieving environmental objectives byreplacing the current price support policieswould require the payment of land and watersubsidies, and the imposition of a fee onfertilizer use (column C, table 1). Undervaluations assumed for the environmentalgoods, water use would expand by about 10.4percent (8.0 percent under trade liberalization)so that the value of ground- water recharge isincreased by 9.5 percent (6.8 percent undertrade liberalization). Fertilizer use would declineby about 2.4 percent (7.6 percent under tradeliberalization). The amount of land in riceproduction would rise by 7.6 percent (4.6percent under trade liberalization), in part tofoster the increase in groundwater recharge.

Table 2 contains a summary of the values for theparameters and the exogenous variables used inthe empirical model. After calibrating thefunctional constants, ki, for each equation basedon the observed data, we find the optimal levelsof the 13 endogenous variables (L, Z, FP, W, Pl,PZ, Pfp, Pw, Qd, P, PP, GW, ME) by solvingequations (1) through (13) simultaneously.

Table 3 provides a summary of key resultsfrom the policy simulations. Columns A to Crelate to simulations performed under theassumption of autarky, and columns D to Freflect limited imports whose allocation iscontrolled by the state trading entity.7

Columns A and D provide the bases forcomparison under the closed economy and limitedtrade options, in that each relates to the ‘currentsupport’ case—one in which the existing systemof price and income support with a land set-asideis in operation. Columns C and F relate to thecase in which price and income supportobjectives are abandoned, being replaced by apolicy in which environmental objectives alone arepursued. To provide a further point of comparison,

7In all simulations where limited imports are allowed under the TRQ, the quota is binding (e.g., imports are equal to the minimumaccess commitment). Since the lower, in-quota tariff rate is set at zero in Taiwan, there are no tariff revenues collected.

17

2. Contrary to what one might expect, thispolicy re-instrumentation would actuallyincrease rice production slightly (0.3 percent)under autarky but would decrease it by only3.5 percent under trade liberalization. Asmentioned above, the relatively smallchanges in rice production are due in largepart to the design of the pre-WTO pricesupport policy. The support price was paidonly on a fixed quantity of production perhectare. Thus, even prior to policy reform, thesupply-inducing price at the margin was thedomestic market-clearing price, rather thanthe support price.

3. The replacement of price and income supportpolicies by environmental policies hassignificant redistributive implications. There isa substantial reduction in transfers toproducers, since the current large paymentsto producers are not needed to achieve theenvironmental benefits and also reduce the

TABLE 2.

Values for the parameters and policy variables of the model for the Taiwanese rice market.

environmental costs associated with riceproduction. The domestic payments fall by93.6 percent (94.2 percent under tradeliberalization).

4. The elimination of income support, in theform of set-aside payments and governmentpurchases of rice, causes imputed land rentsto decline substantially, by an estimated 38.9percent (41.9 percent under tradeliberalization).

5. The pursuit of environmental objectives iswelfare enhancing. Producers lose,consumers gain, and government paymentsare reduced. Net social welfare increasesonce environmental externalities areinternalized in the autarky case, but the gainsare higher under trade liberalization. This isconsistent with the view that tradeliberalization is welfare-enhancing overall, andsuggests that the pursuit of environmentalaims is not inconsistent with freer trade.

18

6. The joint pursuit of redistributive andenvironmental objectives is inferior, in termsof net social welfare, to a policy thatattempts to achieve environmentalobjectives alone. Because themultifunctional non-commodity outputs arenot produced in fixed proportion with output,the compensation of producers associatedwith the income support objective implicitlyvalues groundwater recharge below its

TABLE 3.

Simulations for alternative agricultural, environmental and trade policies on the Taiwanese rice market.

Autarky Limited Trade LiberalizationCurrent Support+ Green Current Support + GreenSupport Env Policy Support Env Policy

(A) (B) (C) (D) (E) (F)

Optimal Environmental PolicyLand compensation (US$/ha) 0.0 -91.2 10.0 0.0 -126.6 10.5

Water compensation (US$/ 1,000 tons) 0.0 1.2 1.2 0.0 1.2 1.1

Fertilizer fee (US$/ton) 0.0 1.7 1.7 0.0 1.7 1.8

Resource Allocation - - - - - - - - - - - - - - - -Percent Change from Base (A) - - - - - - - - - - - - - - - - -Consumption (thousand tons table rice) 1,207.0 -0.1 0.3 0.5 0.3 0.7

Production (thousand tons raw rice) 1,724.0 -0.1 0.3 -3.7 -3.9 -3.5

Land planted to rice (1,000 ha) 332.0 0.0 7.6 -0.4 -0.4 4.6

Labor (1,000 persons) 231.0 0.5 -1.3 -3.9 -3.2 -4.8

Water (107tons) 804.0 11.1 10.4 -2.1 9.8 8.0

Fertilizer (1,000 tons) 829.0 0.3 -2.4 -5.9 -5.2 -7.6

Groundwater recharge 278.0 5.9 9.5 -1.5 4.4 6.8

Methane emissions (103 tons) 35.5 -0.9 2.3 -3.0 -4.3 -1.5

Prices and RevenueImputed land rent (US$/ha) 23.8 -6.5 -38.9 -0.8 -9.9 -41.9

Wages (US$/person) 68.1 0.6 -1.6 -4.9 -4.0 -6.0

Fertilizer Price (US$/ton) 7.0 0.2 -1.2 -2.6 -3.9 -8.6

Water price (US$/1,000 tons) 0.1 41.9 39.1 -6.7 36.6 29.1

Farm price of rice (1,000 US$/ton) 0.5 1.3 -3.2 -5.2 -3.3 -7.3

Farm revenue (108 US$)a 11.3 -1.7 -18.9 -8.9 -25.2 -16.7

Welfare AnalysisConsumer surplus (109US$) 5.8 -0.3 0.6 1.0 0.7 1.5

Producer surplus (109US$) 0.6 -1.9 -21.4 -5.7 -8.9 -27.6

Domestic payments (107US$) 19.9 -5.0 -93.6 5.5 -17.2 -94.2

Groundwater recharge value (107US$) 28 0.3 0.5 -0.1 0.2 0.3

Methane emission value (107US$) 0.3 -0.9 2.3 -3.0 -4.3 -1.5

Social welfare (109US$) 6.5 0.1 1.4 0.7 0.5 2.1

social value and underestimates the socialcost of methane abatement. This is bestseen by the magnitude of the environmentalfees and compensations in the presence ofdomestic distortions created by existingpolicies. Given these distortions,environmental objectives can only beachieved through a land fee, rather than aland compensation, in conjunction with awater compensation.

Notes: aIncludes domestic policy payments.All monitary values are in constant 2001 US dollars. The average NT$/US$ exchange rate in 2001 was 33.8 (GIO, 2004)

19

Sensitivity Analysis

The results in table 3 were derived using themean of the valuation for groundwater recharge(WTP = US$844) and with the cost of methaneemissions associated with the most efficientmanagement strategy (Me = US$86). In order toexamine the robustness of our conclusions, weconducted a sensitivity analysis using theclosed-economy scenario as a point of reference

(table 4). Columns A and C in that tablecorrespond to columns A and C in table 3.Columns G and H give the results obtained usingthe high and low valuations, respectively, fromwillingness to pay studies, as mentioned above.The abatement costs associated with methaneemissions are unchanged from those used ingenerating the results in table 3 for these twosimulations. This table also gives the results thatare obtained when the alternative high (column I)

TABLE 4.

Sensitivity of the policy simulations to alternative values for the environmental variables.

Scenarios With Alternative Environmental Values(A) (C) (G) (H) (I) (J)

Parameter Value

WTP (US$) — 843.8 1,114.1 573.6 843.8 843.8

Me (US$) — 85.5 85.5 85.5 402.4 1,341.9

Optimal Environmental Policy 0.0 0.0 0.0 0.0 0.0

Land compensation (US$/ha) 0 10.0 14.4 5.7 -2.9 -41.1

Water compensation (US$/1,000 tons) 0 1.2 1.4 0.7 1.0 0.8

Fertilizer fee (US$/ton) 0 1.8 1.8 1.8 8.3 27.5

Resource Allocation - - - - - - - - - - - - - Percent Change From Base (A)- - - - - - - - - - - - - -Production (thousand tons raw rice) 1,724.0 0.3 0.3 0.3 0.2 -0.1

Land planted to rice (1,000 ha) 332.0 7.6 7.9 7.4 7 5.1

Labor (1,000 persons) 231.0 -1.3 -1.4 -1.2 -0.8 0.4

Water (107tons) 804.0 10.4 11.9 6.5 9 8

Fertilizer (1,000 tons) 829.0 -2.4 -2.5 -2.2 -3.4 -6.4

Groundwater recharge 278.0 9.5 10.6 6.8 8.3 7.1

Methane emission (103 tons) 35.5 2.3 2.4 2.1 1.5 -0.7

Prices and Revenue

Imputed land rent (US$/ha) 23.8 -38.9 -38.6 -39.1 -39.5 -41.4

Wages (US$/person) 68.1 -1.6 -1.7 -1.5 -1.1 0.5

Fertilizer Price (US$/ton) 7.0 -1.2 -1.3 -1.1 -1.7 -3.3

Water price (US$/1,000 tons) 0.1 39.1 45.5 23.2 33.3 29.4

Farm price of rice (1,000 US$/ton) 0.5 -3.2 -3.4 -2.9 -2.1 1.1

Farm revenue (108 US$)a 11.3 -18.9 -18.7 -19.2 -19.1 -19.3

Welfare Analysis

Consumer surplus (109US$) 5.8 0.6 0.7 0.6 0.4 -0.2

Producer surplus (109US$) 0.6 -21.4 -21.1 -21.9 -21.6 -21.7

Domestic payments (107US$) 19.9 -93.6 -92 -96.5 -99.4 -114.3

Groundwater recharge value (107US$) 28 0.5 32.7 -31.8 0.4 0.4

Methane emission value (107US$) 0.3 2.3 2.4 2.1 377.5 1458.1

Social welfare (109US$) 6.5 1.4 2.8 0 1.2 0.6

Notes: aIncludes domestic policy payments.All monitary values are in constant 2001 US dollars. The average NT$/US$ exchange rate in 2001 was 33.8 (GIO, 2004)

20

and very high (column J) abatement costsassociated with alternative managementstrategies are assumed (Boisvert et al. 2004)).The valuation attached to groundwater recharge isthat used in deriving the results in table 3. Thekey points from the sensitivity analysis are:

1. As might be expected, a lower valuation ongroundwater recharge results in a reduction inthe land and water subsidies required toachieve environmental objectives; a highervaluation results in larger compensations.

2. As the economic costs associated withmethane emissions increase, the optimal feeon fertilizer use rises, and land compensationis replaced by a land fee. These measuresresult in a reduction in both the amount ofland under rice cultivation and the productionof rice.

3. Although the net effect varies, social welfareis increased by the pursuit of environmentalobjectives in comparison to current policies.The substantial redistributive impact of thechange in policy aims (as evidenced by thereduction in producers’ surplus and landrents) remains.

Distributional Consequences of PolicyChange

From these results, there are gains in social welfarefrom the re-instrumentation of the domestic ricepolicy in Taiwan. The efficiency gains from replacingprice supports and land set-aside payments withpolicies aimed directly at providing optimal levels ofmultifunctional output and limited trade liberalizationare distributed toward consumers and away fromfarm households producing rice. Therefore, tofacilitate the process of adjustment to this newpolicy regime, there may be political reasons toaddress these distributional issues. Importantthough they may be, we have emphasized abovethat these distributional issues fall outside ourdefinition of multifunctionality and, therefore, they

must be addressed with a different policy approach.The objectives of a redistribution policy would be toensure a smooth transition to a new long-termequilibrium number of rice farmers of appropriatesize, and to address issues related to the economicviability of the affected farm households.

In some countries, policy strategies, such ascompensatory payments to producers, have beenused to ease the adjustment process resultingfrom the reform of domestic agricultural policies.Recently, Australia made payments to dairyfarmers as part of the reform of domestic dairypolicies (Harris and Rae 2006). Recognizing thattrade liberalization and a reduction in traditionalcommodity support can reduce asset values inagriculture, steps have been taken recently byother countries to provide compensation to assetowners for such reductions. In the United States,for example, such payments were made to formerholders of marketing quotas for peanuts as partof the introduction of a more market-orientedgovernment support program for that commodityunder the 2002 Farm Act (Dohlman et al. 2004).

To gain some perspective on the implications ofproviding adjustment assistance, we assume herethat Taiwanese rice farmers are compensated forlosses in land values due to the reduction in pricesupports and land set-aside payments. Focusing onthe scenario from table 3 for the case wheretraditional policies are replaced with themultifunctional policies combined with tradeliberalization, current annual returns to land areestimated to fall by nearly 42 percent (e.g., thechange in imputed land rent, table 3). Tocompensate for this annual loss in land value, directpayments of about US$337/ha. would be required.The total cost, if this payment is made only to landin production (including the 4.6 percent increase inrice area) would be US$117 million, or 59 percent ofthe annual government outlays for price supportsand land set-aside payments. If direct payments ofthis size were also made to the remaining set-asideland, the cost would rise to US$158 million or 79percent of the budget outlays under the traditionalsupport policies.8

8So as to not affect production, payments to this remaining set-aside land may well be necessary for the direct payments to meetconditions for inclusion in the WTO ‘green box’ provisions.

21

While a one-time payment of this size wouldcompensate producers for a single year’sreduction in returns to land, it would not accountfor losses in future years. To account for lossesin future years, the government could make amuch larger one-time payment, or make annual

payments for a fixed number of years into thefuture. The one-time payment for the lost valueof quota permits in the United States, forexample, was equal to the loss in valuediscounted at 5 percent over a 24-year period(Orden 2005).

Policy Implications

The approach developed in this reportdemonstrates how optimal policies for the supplyof non-commodity (environmental) attributesassociated with Taiwanese rice production can bedetermined. We have shown that a policyaimed at securing an appropriate supply of amajor positive attribute (groundwater recharge)and containing a negative attribute (methaneemissions) would require policyre-instrumentation. Fees and compensations forthe use of inputs used in rice production wouldneed to replace the price and income supportmeasures that are currently used. We have alsoshown that the modest liberalization of trade thathas taken place since Taiwan joined the WorldTrade Organization, is not inconsistent withachieving key environmental objectives.Recognizing that the specific numerical resultsare sensitive to parameter values, we havedemonstrated the robustness of the qualitativeconclusions on appropriate policy design undersubstantial variation in the valuation parametersfor environmental attributes.

Our results also have direct policyimplications for the distribution of water betweenagricultural and industrial uses in Taiwan. Whencurrent domestic support is replaced by theoptimal environmental policies, the demand ofwater in rice production increases, reflecting itsvalue both in rice production and in producing avaluable non-commodity output of groundwaterrecharge. To put it differently, while the currentdomestic price support and set-aside policy do

provide income support for agriculturalproducers, our analysis shows that the currentpolicies understate the true (private plus social)value of water used in agricultural production. Itis only through policies such as those examinedhere that one can be assured that importantscarce natural resources can be allocatedoptimally between important sectors of theTaiwanese economy.

Although the results obtained from this singleempirical application may not be generallyapplicable to other countries and agriculturalsystems, we believe that they shed some light onthree important issues surrounding the debate onmultifunctionality.

The first is that a policy approach that treatsnon-commodity attributes in agriculture assecondary objectives to existing aims, such asincome support, may well result in sub-optimaloutcomes. As the Taiwan case demonstrates, itdoes not follow that the pursuit of income supportgoals through the use of traditional policymeasures will result in the desired supply ofenvironmental goods. It may be possible toachieve that supply at far lower costs, both to thegovernment and to society as a whole, if policyinstruments are employed that are more suited forachieving environmental goals than the existingmeasures. If it is indeed the case, as somewould argue, that agriculture’s role in the supplyof environmental goods is an increasinglyimportant issue, it may not be appropriate orsufficient to make marginal adjustments in the

22

settings of existing agricultural policy instrumentsto achieve the desired outcome; a more radicalredesign of policy may be required.

The second issue illuminated by our analysisis that this more radical redesign of policy mayhave significant redistributive implications,particularly through reductions in farm assetvalues, that may have to be addressed for politicaland equity reasons and to help control farmers’entry and exit from agriculture. Our results indicatethat the efficiency gains giving rise to increasedsocial welfare, and the direct payments needed tocompensate farmers for loss in annual returns toland can be achieved at a substantially lower costto the government than the cost associated withtraditional farm support policies.

The third issue illuminated by our analysis isthat agricultural trade liberalization may not beinconsistent with the pursuit of domesticenvironmental objectives. Put differently, ourresults provide empirical evidence supporting theearly arguments by Bhagwati and Ramaswami(1963), as well as others, that trade policy shouldnot be used to correct domestic distortions. Inthe light of our findings, the distortions created bythe lack of markets for the positive and negativeexternalities associated with agriculturalproduction are best addressed directly throughdomestic fees and compensations for the use ofinputs, while at the same time allowingconsumers to benefit from lower product pricesresulting from reduced import protection. The

maintenance of high product prices through importprotection seems an inefficient way to achieveenvironmental aims, and could run counter toachieving those aims.

In conclusion, the results of this empiricalinvestigation into the effects of there-instrumentation of agricultural policy in Taiwanare extremely encouraging. They demonstrate thatpolicy reform can contribute to ensure optimallevels of multifunctional outputs, and that it alsocan address non-trade concerns that fall outside aworkable definition of multifunctionality. Directionsfor additional research are also suggested in theefforts made to reconcile existing social values orcosts of the multifunctional attributes of rice. Asis well understood in the environmental economicpolicy arena, many of the multifunctional attributesdepend on local soil and water conditions as wellas other factors that affect agricultural production.The values placed on these attributes can differtoo. Therefore, there is need to extend thisresearch by considering explicitly these regionaldifferences. If they prove to be significant, thenpolicy re-instrumentation may have to beimplemented at the regional or local level. Inaddition, these types of policy analyses need tobe extended to include additional multifunctionalattributes. To the extent that policies recognizethe contribution of farm production to a broaderrange of multifunctional attributes, the distributionalconsequences of policy reform may be diminished,thus reducing the need for adjustment assistance.

23

Literature Cited

Abbott, P.; Paarlberg, P. 1998. Tariff rate quotas: Structural and stability impacts in growing markets. AgriculturalEconomics 19 (3): 257-267.

ADB (Asia Development Bank). 1998. Asia least cost greenhouse gas abatement strategy. Manila, The Philippines:People’s Republic of China. Global Environment Facility.

Asian Productivity Organization (APO). 2001. Role of multifuctionality in agricultural policy reforms. Tokyo, Japan: APO.

Barends, F.; Brouwer, F.; Schroeder, F.; eds. 1995. Land subsidence: By fluid withdrawal, by solid extraction, theoryand modelling, environmental effects and remedial measures. Proceedings of the Fifth International Symposiumon Land Subsidence, The Hague, the Netherlands.

Barnett, A.H. 1980. The Pigouvian tax rule under monopoly. American Economic Review 70 (5): 1037-1041.

Baumol, W.; Oates, W.E. 1988. The theory of environmental policy, 2nd Ed., Cambridge, MA: Cambridge UniversityPress.

Bhagwati, J.; Ramaswami, K. 1963. Domestic distortions, tariffs, and the theory of optimal subsidy. Journal of PoliticalEconomy 71 (1): 44-50.

Blandford, D.; Boisvert, R.N. 2002. Multifunctional agriculture and domestic/international policy choice. Estey Journalof International Law and Trade Policy 3 (1):106-118.

Blandford, D.; Boisvert, R.; Fulponi, L. 2003. Non-trade concerns: Reconciling domestic policy objectives with freertrade in agricultural products. American Journal of Agricultural Economics 85 (3): 668-673.

Blandford, D.; Boisvert, R.N. 2005. Non-trade concerns: Reconciling domestic policy objectives with trade liberalization.International Journal of Agricultural Resources, Governance and Ecology 4 (3-4): 277-291.

Boisvert, R. N. 2001a. A note on the concept of jointness in production. Annex 1. In Multifunctionality: Toward ananalytical framework. Paris, France: Organization for Economic Cooperation and Development OECD.

Boisvert, R. N. 2001b. Joint production in four outputs: Two commodities and positive and negative externalities.Annex 2. In Multifunctionality: Toward an analytical framework. Paris, France: Organization for EconomicCooperation and Development OECD.

Boisvert, R. N.; Chang, H.H.; Barker, R.; Levine, G.; Matsuno, Y.; Molden, D. 2003. Refining the positive and negativeexternalities of Taiwanese rice production. Paper presented at the ICID Asian Regional Workshop, Taipei, Taiwan,November 10-12, 2003.

Boisvert, R. N.; Chang, H.H.; Barker, R.; Levine, G.; Matsuno, Y.; Molden, D. 2004. Refining the positive and negativeexternalities of Taiwanese paddy-rice production. Annex III. In Accounting for agricultural and nonagriculturalimpacts of irrigation and drainage systems. A Report of Research in Taiwan and Sri Lanka in 2003. Working Paper68 Colombo, Sri Lanka: International Water Management Institute.

Boughner, D.; de Gorter, H. 1999. U.S. dairy policy and the agreement on agriculture in the WTO. Canadian Journalof Agricultural Economics 47(5): 31-42.

Carson, R.; Flores, N.E.; Hanemann, W.M. 1998. Sequencing and valuing public goods. Journal of EnvironmentalEconomics and Management 36 (3): 314-323.

Chang, H.H. 1998. Bounded price variation model with rational expectation and price risk-an analysis for the pricepolicy and simulation in Taiwan rice market. Unpublished Master’s Thesis, National Taiwan University.

Chang, H. H.; Boisvert, R.N.; Blandford, D. 2005. Achieving environmental objectives under reduced domesticagricultural support and trade liberalization: An empirical application to Taiwan. Agricultural and ResourceEconomics Review 34 (1):16-31.

Chen, M.C. 2001. Evaluation of environmental services of agriculture in Taiwan. Paper presented at the InternationalSeminar on Multi-functionality of Agriculture. Japan.

Chen, Y.H.; Chang, H.H. 2002. The impact of Taiwanese rice market under the minimum access quota and directpayment policies. Journal of Agricultural Finance 47: 95-111. (in Chinese)

24

Chen, J.; Wu, S.; Chang, K. 2002. The measurement of externalities of paddy field: An application of contingentvaluation in Taiwan. Annex 3. In Accounting of agricultural and nonagricultural impacts of irrigation and drainagesystems: A study of multifunctionality in rice, eds. Y. Matsuno, H. Ko, C. Tan, R. Barker and G. Levine. WorkingPaper 43. Colombo, Sri Lanka: International Water Management Institute.

CoA (Council of Agriculture). 1997. The 100-year history of Taiwanese food policy. Taipei Taiwan. (in Chinese)

CoA (Council of Agriculture). 2002. Rice paddy utilization adjustment program. Taipei, Taiwan. (in Chinese)

Dohlman, E.; Hoffman, L.; Young, E.; McBride, W. 2004. Peanut policy change and adjustment under the 2002 farmact. Washington, D.C.: Economic Research Service (USDA). Outlook Report No. (OCS04G01) 32. July.

Floyd, J. 1965. The effects of farm price supports on the return to land and labor in agriculture. Journal of PoliticalEconomy 73 (2):148-158.

Fraser, R. 2003. An evaluation of the compensation required by European Union cereal growers to accept theremoval of price support. Journal of Agricultural Economics 54 (3): 431-445.

Fu, T.T.; Shei, S.Y. 1996. Agriculture as the foundation for development: The Taiwanese story. Paper presented atConference in Memory of Professors Ta-chung Liu and Sho-Chieh Tsiang on Government and Markets: TheRelevance of the Taiwanese Performance (1945-1995) to Development Theory and Policy, Cornell University.

Fu, T.T.; Chen, B. 1995. The political economy of Taiwanese rice market and welfare analysis. Journal of TaiwaneseBank 46 (1): 304-325. (in Chinese)

Gardner, B. 1987. The economics of agricultural policies. New York, Macmillan.

Gisser, M. 1971. A model for agricultural policy—an addendum. American Journal of Agricultural Economics53 (1):125-128.

Government Information Office. 2004. Taiwan yearbook 2004. Taipei, Taiwan.

Griffin, R.; Bromley, D. 1982. Agricultural runoff as a nonpoint externality: A theoretical development. AmericanJournal of Agricultural Economics 64 (3): 547-552.

Harris, D.; Rae, A. 2006. Agricultural policy reform and industry adjustment in Australia and New Zealand. In Policyreform and adjustment in the agricultural sectors of developed countries, eds. D. Blandford and B. Hill. Wallingford,Oxfordshire: CABI Publishing.

Hoehn, J. P.; Randall, A. 1989. Too many proposals pass the benefit cost test. American Economic Review79 (3):544-551.

Holtermann, S. 1976. Alternative tax systems to correct for externalities, and the efficiency of paying compensation.Economica 43:1-16.

Huang, C.H. 1992. Effects of government program on rice acreage decisions under rational expectations-the case ofTaiwan. American Journal of Agricultural Economics 74 (2): 310-317.

Huang, S. 2001. Taiwan’s rice import market to open with WTO accession. Rice situation and outlook yearbook /RCD-2001, Washington, D.C.: Economic Research Service. US Department of Agriculture.

Hyman, R. 2001. A more cost-effective strategy for reducing green house gas emissions: Modeling the impact of methaneabatement opportunities. Unpublished Masters Thesis, Massachusetts Institute of Technology, Cambridge, MA.

Kolstad, C. 2000. Environmental economics. New York: Oxford University Press.

Krutilla, K. 1991. Environmental regulation in an open economy. Journal of Environmental Economics and Management20 (2):127-142.

Leathers, H. 1991. Allocable fixed inputs as a cause of joint production: A cost function approach. American Journalof Agricultural Economics 73 (4):1083-1090.

Lee, D. 1975. Efficiency of pollution taxation and market structure. Journal of Environmental Economics and Management2 (1): 69-72.

25

Lichtenberg, E.; Zilberman, D. 1986. The welfare economics of price support in U.S. agriculture. American EconomicReview 76 (5):1135-1141.

Lin, J.H. 2000. The rice farmer decision impact by agricultural policy. Unpublished Masters Thesis. National TaiwanUniversity, Taiwan. (in Chinese)

Lin, S.; Pon, P.; Hsu, S. 2002. Baseline methane and nitrous oxide emission estimation and forecasting of Taiwaneseagricultural sector. Paper presented at the 1st Applied Economics Annual Conference, National Chung-HsingUniversity, Taichung, Taiwan. (in Chinese).

Maier, L. 1991. The costs and benefits of U.S. agricultural policies with imperfect competition in food manufacturing.Unpublished PhD Dissertation. Cornell University, Ithaca, NY.

Matsuno, Y.; Ko, H.; Tan, C.; Barker, R.; Levine, G. 2002. Accounting of agricultural and nonagricultural impacts ofirrigation and drainage systems: A study of multifunctionality in rice. Working Paper 43. Colombo, Sri Lanka:International Water Management Institute.

Ohba, Y. 2001. An economic analysis of acreage diversion and the tariff-rate import quota for rice in Japan. UnpublishedMasters Thesis. Cornell University, Ithaca, NY.

Ohnishi, R.; Nakanishi, N. 2001. Water conservation function and appropriate management of agricultural land.Proceedings of the International Seminar on Multi-functionality of Agriculture, Japan, October 17-19, 2001.

Orden, D. 2005. Key issues for the next farm bill: Is a farm program buyout possible? Paper presented at theAgricultural Outlook Forum, Washington, D.C.: February 24, 2005.

OECD (Organization for Economic Cooperation and Development). 2001. Multifunctionality: Towards an analyticalframework. Paris, France : OECD.

Peterson, J.; Boisvert, R.; de Gorter, H. 2002. Environmental policies for a multifunctional agricultural sector in openeconomies. European Review of Agricultural Economics 29 (4): 423-443.

Plott, C. 1966. Externalities and corrective taxes. Economica (New Series) 33: 84-87.

Randall, A. 2002. Valuing the outputs of multifunctional agriculture. European Review of Agricultural Economics29 (3): 289-307.

Romstad, E. 2004. Multifunctionality — focus and resource allocation. In Multifunctional agriculture, policies andmarkets. Understanding the critical linkages. Proceedings, Part 1, 90th EAAE Seminar, Rennes, France, October28-29, 2004.

Rosine, J.; Helmberger, P. 1974. A neoclassical analysis of the US farm sector: 1948-1970. American Journal ofAgricultural Economics 56 (4): 717-729.

Rude, J. 2001. Multifunctionality: An examination of the issues and remedies. In Globalization and agricultural tradepolicy, eds. H.J. Michelmann, J. Rude, J. Stabler and G. Storey. Boulder, CO: Lynne Rienner Publishers.

Shortle, J.R.; Horan, R.D.; Abler, D. 1998. Research issues in nonpoint pollution control. Environmental and ResourceEconomics Review 11 (3-4): 571-585.

Shortle, J.; Dunn, J. 1986. The relative efficiency of agricultural source water pollution control policies. AmericanJournal of Agricultural Economics 68 (3): 668-778.

Shumway, C.; Pope, R.; Nash, E. 1984. Allocable fixed inputs and jointness in agricultural production: Implications foreconomic modeling. American Journal of Agricultural Economics 66 (1):72-78.

Sophia H. 2001. Taiwan’s rice import market to open with WTO accession. Rice Situation and Outlook Yearbook,OECD.

Sophia H.1993. Structural change in Taiwan’s agricultural economy. Economic Development and Cultural Change42 (1): 43-65.

Tirole, J. 1988. The theory of industrial organization. Cambridge, MA: MIT Press.

Tsai, M. 1993. The public benefit externality of Taiwanese paddy rice. Taipei, Taiwan: Council of Agriculture.

26

Tsai, S.; Wann, J. 1995. Selecting functional form for empirical economic model – the case of Taiwan rice productionfunction. Taiwanese Agricultural Economic Review 1 (2):187-216. (in Chinese with English Summary)

Tyrchniewicz, E.; Schuh, G. 1969. Econometric analysis of the agricultural labor market. American Journal of AgriculturalEconomics 51 (4): 770-787.

U.S.E.P.A (U. S. Environmental Protection Agency). 1999. U.S. methane emission 1990-2020: inventories, projections,and opportunities for reductions. Washington, D.C.: Office of Air and Radiation, September, 1999.

Yang, M.C. 1993. The political economy of rice policy in Taiwan. Unpublished PhD. Dissertation. National TaiwanUniversity, Taiwan. (in Chinese).

Yang, S. 2000. The green gas emission estimation of agriculture, livestock sectors- project final report. EPA-89-FALL-03-145, Taipei, Taiwan: National Taiwan University. (in Chinese).

Yang, M.; Chen, H. 2000. A policy choice of rice import Issue in Taiwan. In Food security in asia: Economics andpolicies, eds. W. Chern, C. Carter and S. Shei. Northampton, MA: Edward Elgar Publishing Limited.

27

Annex A

Brief History of Taiwanese Rice Policy

Introduction

Rice is the staple crop in many Asian countries,especially Japan, Korea, and Taiwan. AfterTaiwan separated from Mainland China in 1946,the Taiwanese government began to intervene inthe rice market both to ensure an ample foodsupply and for reasons of food security. As theeconomy has developed since that time,government policy has also evolved in responseto different economic circumstances. Between1946 and 1997, we can roughly divide thisevolution of government rice policy into fourperiods (Fu and Chen 1995; CoA 1997). Duringthe first two periods, encompassing the years1946-1973, the goals of government rice policywere to stabilize prices and to help expandagricultural supply. During this time, theagricultural sector also supported other economicactivity by supplying vital resources to the rest ofthe economy. During the latter two periods,encompassing the years 1974-1997, the goals ofgovernment policy shifted dramatically—theprimary concern focused on increasing farmers’incomes through the adoption of a price supportprogram and other policies. By the end of thetwentieth century, agriculture had become aprotected sector, supported by resources fromother sectors of the general economy (Fu andShei 1996).

Since 1998, and as a condition for becominga member of the World Trade Organization(WTO), the Taiwanese government has had toreexamine the goals of its agricultural policy,effectively entering into the fifth period in thehistorical evolution of rice policy. In particular, tomeet the free trade objectives of the WTO, thegovernment’s efforts to support its domesticagricultural sector will have to rely less on pricesupports and other domestic policies that distort

international trade. Furthermore, along withdecreasing domestic support levels, thegovernment must also open agricultural marketsto increased international trade.

These policy changes will have dramaticeffects on the Taiwanese rice market and on theincomes of farmers. As Taiwan’s agriculturaleconomy is subjected to these significant internaland external pressures, government policydecision makers will be preoccupied with theissues surrounding structural policy adjustmentfor years to come. How will Taiwan’s agricultureof small family farms survive? What are thebarriers to future structural adjustments? What isthe appropriate new direction of Taiwan’s futureagricultural policy after 1998?

To place these issues in proper perspective,the purpose of this annexure is to recapitulate thecourse of Taiwan’s structural change in agriculturesince 1946, including the discussion of theevolution of main agricultural policies. The historyof agricultural policy initiatives is summarized intable A.1.

A Period to Encourage Production andControl of Rice Supply (1946-1968)

In 1946, immediately following the internal warbetween Taiwan and Mainland China, there was aserious shortage of manufacturing facilities tosupport the Taiwanese economy. The generalprice level was also high, caused by the scarcityof materials and food, including rice. Notsurprisingly, the government was preoccupied withproviding enough food and basic materials tosupport the national population and stabilize thegeneral price level.

To increase the aggregate food supply, theTaiwanese government implemented a series of

28

land reform policies to provide economic incentivesfor rice farming. These included policies such asthe ‘37.5 percent farm land rent reduction program’in 1949, the ‘sales of public farm land policy’ in1951, and the ‘Tiller Act’ in 1953. The governmentprovided farm inputs, such as fertilizer andtechnical support for farmers. The government alsofacilitated the construction of irrigation systems.As a result of these policies, the number of ricefarmers increased by one third, and the foodsupply increased rapidly as well.

To effectively control the food supply, policiesknown as the ‘Taiwan Land Tax with RiceQuantity’ and the ‘Rice Fertilizer Barter’ programswere also launched in 1946. Through thesecombined policies, rice farmers were allowed topay the land tax with a portion of their riceproduction, and in turn, also received a fertilizerallocation. At that time, the government alsorestricted the numbers of private food stores thatcould sell rice to the general public.

The combined effect of these demand- andsupply-side policies was a rapid increase in riceproduction, which helped to support the rest ofthe general economy and stabilize the generalprice level. Between 1946 and 1966, for example,rice production increased from 894 thousand MTto 2,379 thousand MT (an annual growth rate of2.7 percent). Although these kinds of policieswere effective in increasing the food supply, thegovernment actually taxed rice farmers by forcingthem to exchange production inputs at priceslower than the market price (Lin 2000). Underthese policies, farmers had to increase riceproduction to survive financially.

Self-sufficiency and a Period of ExportDifficulties (1969-1973)

During the ‘Green Revolution’ of the late 1960sand into the 1970s, world supplies of riceoutstripped world demand, due primarily to thedevelopment and widespread adoption of highyielding varieties. Therefore, the world rice pricedropped dramatically. The problems facingTaiwanese rice farmers became particularly acute

as the major importers of Taiwanese rice, suchas Japan, were among the widespread adoptersof new varieties to promote food self-sufficiencyand decrease their dependence of imported rice.The loss of these international markets for riceled to serious reductions in Taiwan’s foreignexchange earnings.

Due to the rapid growth in Taiwan’s industrialeconomy during this same period, agriculture’ssignificance in the overall economy began todecline. Reflecting this structural change,agriculture’s share of GNP had fallen to justunder 16 percent by 1969. In response to thesepervasive economic forces, the governmentreoriented its agricultural policy toward improvingfarmers’ incomes and reducing costs ofproduction. To achieve these objectives andreduce farmers’ tax burdens, some policies, suchas the ‘Rice Fertilizer Barter’ program, wereabolished. By doing so, the government also lostmuch of its control over rice production.

A Period to Support Prices andStabilize Food Supply (1974-1983)

The world energy crisis, beginning in the mid-70s,was accompanied by a serious worldwide shortfallin rice production in 1973. By 1975, the worldprice of rice had risen substantially, and theTaiwanese government’s stocks of rice werenearly depleted. In 1974, the governmentestablished the ‘Food Stabilization Fund’ to regaincontrol of rice production in the face of theseshocks and to stabilize the price at a reasonablelevel. With a budget of US$89 million, thepurpose of the fund was to purchase rice fromfarmers at prices exceeding world market prices.Between 1974 and 1976, the government placedno limit on its purchases. Rice productionincreased rapidly from 2,254 thousand MT in1973 to its historical peak of 2,713 thousand MTin 1976.

By this time, the persistent rice surplus wasbeginning to put heavy pressure on government’sstorage capacity and budget outlays. To balancethe budget and reduce government stocks of rice,

29

the government adjusted the price supportprogram by limiting planned purchases to 970 kgper hectare. To provide additional assistance tofarmers, the government also launched the‘Guidance Purchase Program’ in 1978, wherebyfarmers’ associations in each county could alsopurchase rice from local farmers. Beginning inthat year, rice farmers effectively had two waysto participate in the government’s support priceprogram (Yang 1993).

Rice Diversion and a Period ofDemand-Supply Balance (1984-1998)

The government’s support program for rice duringthe 1970s and early 1980s clearly protectedfarmers’ incomes and controlled market supply,and it marked a significant change in thehistorical development of the government’s ricepolicy (Sophia 1993). The government paid aheavy price (particularly in terms of budgetoutlays) to reach these policy goals.

With continued economic growth andincreases in personal income during the 1980s,household consumption patterns (including foodconsumption) became more diverse. As percapita rice consumption began to decline, therewas an ever-increasing excess supply of rice. Inorder to balance the demand and supply in thedomestic market for rice and reduce the budgetoutlays for purchasing rice, the governmentlaunched the first 6 years of its ‘Rice DiversionProgram’ in 1984.

Under the ‘Rice Diversion Program’, ricefarmers were encouraged to plant high-valuedgrains and horticulture crops on land previouslyused exclusively for rice production. From 1984to 1987, the incentive was an in-kind payment of1.5 MT of rice per hectare converted to theseother uses. Between 1988 and 1996, the in-kindpayments were replaced by direct monetarypayments. Unfortunately, these efforts met withlittle success, in part because the economicincentives were not sufficiently large (Huang1992). There were other reasons for why ricefarmers chose to keep land in rice production.First, and perhaps foremost, the price support

program could still offer substantial incomeprotection. Furthermore, farmers could stillproduce two crops of rice a year even in thenorthern-most counties of Taiwan. Finally,irrigation systems currently in place weredesigned primarily for rice production, thus,making rice less susceptible than other crops todrought and other natural disasters (Sophia 2001).By 1989, the government had recognized thelimited success of these new policy initiatives,and it decided to continue its ‘Rice DiversionProgram’. This program continued until 1997.

Joining WTO: Decreased Supply andGovernment Support (1998-present)

Since 1998, the Taiwanese government has hadto reexamine the goals of its agricultural policy asa condition for becoming a member of the WorldTrade Organization (WTO). To meet the free tradeobjectives of the WTO, the government had toagree to reduce domestic price supports andother domestic policies that distort internationaltrade. Furthermore, along with decreasingdomestic support levels, the government had toopen agricultural markets to increasedinternational trade. Thus, in discussingcontemporary agricultural policy, we must dealwith both domestic and trade policy.

Domestic Policy

To respond to the WTO requirements fordecreasing the levels of domestic support forrice, the Taiwanese government launched a4-year program called the ‘Rice Paddy UtilizationAdjustment Program’ (RPUAP) in 1998. Underthis program, rice farmers were paid a directpayment to set their land aside. Alternatively,they could also receive direct payments (but indifferent amounts) to plant green manure crops,or to rotate the rice crops with other crops (tableA.2). To decrease further the Aggregate Measureof Support (AMS), the government also abolishedthe price support program for other grains in thesame year.

30

Currently, the government is operating undera new 4-year RPUAP, launched in 2001. Thespecial feature of this program was to enlargethe fallow area through higher direct paymentsthan were allowed during the first RPUAP. Thegovernment’s planned set-aside area is 32,000hectares, and based on current yield estimates,the set aside of this area would reducedomestic production by an amount exactlyequal to the amount of rice that Taiwan hasagreed to import from other WTO members(144,720 MT). During this current 4 years of theRPUAP, the government also announced thatfor political and other reasons, it would alsokeep the current price support program. Thus,there are two remaining government policiesaffecting the domestic market for rice: a) pricesupports; and, b) a set-aside program withdirect payments.

Trade Policy

In January 1998, Taiwan became an officialmember of WTO after 142-member governmentsformally endorsed the accession package. By theend of 2002, Taiwan had to allow rice importsthrough a minimum market access quota of anamount equal to 8 percent of domesticconsumption during the base period, 1990-1992.This quota is set at 144,720 MT on a brown ricebasis.9 Although this condition is similar to whatwas required of Japan when it joined the WTO in1995, the Taiwanese government has alsopromised that the imported rice from other WTOmembers would not be used for food aid or asexports to other countries. Hence, this affirmationforces imported rice to be sold in the domesticmarket, thus competing directly with domesticallyproduced rice.

In return, Taiwan is allowed to control 65percent of total imported rice with an additionalmarkup of up to US$0.70/kg when it enters the

9According to the CoA, the minimum access quota for the following years will be decided during the new WTO negotiations.

country. Private traders account for the remaining35 percent (Sophia 2001; Lin 2000). For thisreason, Taiwan’s government still has somepower to control the rice market, but itseffectiveness is diminished substantially fromwhat it was under previous policy regimes.

Challenge for Current Rice Policy

In examining these changes in Taiwanese ricepolicy, it is important to note that the set-asidepolicy is different from the decoupled directpayment system adopted by the United States intheir most recent farm legislation. Although nottrue of the U.S. decoupled payments, farmers inTaiwan must fallow some area to receive thedirect payment. Accordingly, this regulation maystill have some effect on farmers’ productiondecisions, and it is an open question as towhether or not these payments will be classifiedas ‘green box’ by the WTO. Furthermore, someview the current set-aside policy as just ashort-term adjustment policy to meet the 20percent reduction in AMS required by the WTO(Chen and Chang 2002).

The minimum access quotas for rice importsfor future years will be determined during the nextround of WTO negotiations. There are perhapsadvantages to minimum access quotas, one ofthem being that the domestic rice market can beaffected by imported rice, not by the world ricemarket. This may contribute to stabilizing thedomestic price of rice. But in 1999, Japanchanged their rice import policy to a tariff-quotasystem. Thus, Taiwan may be disadvantaged inthe next round of negotiations if it still insists onadhering to the minimum access quota system. IfTaiwan follows Japan in adopting a tariff-ratequota system, the domestic price of rice will beset by the world price, and the government willnot be able to exert much influence on thedomestic rice market.

31

TABLE A 1.

A history of Taiwanese rice policy since 1946.

Period Year Policy

1946-1968 1946 Control number of Food Stores

Taiwan Land Tax with Rice Quantity & Rice Fertilizer Barter Policies

1947 The Sales of Public Farmland Policy

1949 37.5 percent Farm Land Rent Reduction Program

1950 Quantity of rice production reached Pre-World War II level, government started toexport rice

1951 Legislative Yuan agreed to 37.5% Farmland Rent Reduction Program

Executive Yuan agreed to The Sales of Public Farmland Policy

1953 The Tiller Act program was enforced

1960 Implement the farmland reforms

1966 Implement Ten-year Economic Construction Program

1969-1973 1970 Program to increase miscellaneous grains

Japan prohibited foreign rice imported

1972 The Executive Yuan abolished Rice Fertilizer Barter program

1973 The Executive Yuan decided not to export rice and fertilizer

1974-1983 1974 World energy crisis began

The Executive Yuan allotted funds for the Food Stabilization Fund

Implemented the Price Support Program with unlimited purchase

1976 Quantity of brown rice was 2.71 million MT, the historical peak

1977 Price support program has changed to planned purchase with 970 kg of rice/ha. since1977

1978 Implemented another program, Guidance Purchase, to price support program.

From then on, the support of rice excess market demand

1983 Abolished the “Food Store” system

1984-1997 1984 Implemented the first six years of the Rice Land Diversion Program to decrease thesupply

1987 Abolished the Taiwan Land Tax with Rice Quantity Program

1989 Implemented the second six years of the Rice Land Diversion Program

1995 The second six years of Rice Land Diversion Program supposed to end

The Executive Yuan agreed to continue it through June of 1997

1996 The Executive Yuan announced first four-year Rice Paddy Utilization

Adjustment Program, which was implemented in January of 1998

1998-2002 1998 Abolished Price Support Program for Misc. Grains

Implemented four-year Rice Paddy Utilization Adjustment Program

Import U.S. rice through a Minimum Access Quota— WTO negotiations

2001 The Executive Yuan agreed to continue Rice Paddy Utilization

Adjustment Program, chose low quality and quantity county set aside

2002 Rice Paddy Utilization Adjustment Program; increase direct payment

Source: Council of Agriculture (2002), Chang (1998), and Fu and Chen (1995)

32

TABLE A 2.

Direct payments associated with the RPUAP in 2002.

Direct Payment US$/ha

Item Special Land Set-aside Land Plant Green Manure Crops

Basic Payment 799 799 799

Group Rewards 59 59

Additional Rewards 148 296

(administration fee)

Upper Bound on Payment 799 1,006

1998 1,213

2002 1,361

Source: CoA (1997 and 2002)

Note: Monitary values are in constant 2001 US dollars. The average NT$/US$

exchange rate in 2001 was 33.8 (GIO 2004)

33

0),;,,( 21 =ZLEEqT

Annex B

The Mathematical Formulation of a Market Model for TaiwaneseRice Incorporating Multifunctional Policy Design

10In the empirical application above, the two jointly produced non-commodity outputs, E1 and E2, are positive and negative environmentalexternalities, respectively. Thus, throughout the report, the terms non-commodity outputs, multifunctional outputs, and environmentalexternalities are used interchangeably.

11The case of non-allocable factors could also be discussed within the context of technical interdependencies. See Shumway et al.(1984) and Leathers (1991) for rigorous mathematical definitions of jointness in production.

12For example, to compensate for the fact that the level of a negative environmental externality under imperfect competition is alreadybelow its level under perfect competition, the socially optimal tax must be set at less than the marginal social damage (e.g., Lee 1975;Barnett 1980). Lichtenberg and Zilberman (1986) examine the welfare impacts of revenue support programs (e.g., price support,marketing orders, and import quotas) in agricultural product markets.

Introduction

We begin to formulate a conceptual model of theTaiwanese rice sector by assuming fixed resourceendowments of two inputs (L, Z), and byspecifying a general transformation function forboth rice production (q), and two non-commodityenvironmental outputs (E1 and E2). It can berepresented as:

In addition to the normal regularityproperties, the representation of multifunctionalagriculture requires that commodity andnon-commodity outputs are produced jointly, andthe non-commodity outputs must be ‘public’goods or technical externalities that are notvalued in organized markets.10 In this case, it iseasiest to think of the joint production in whichthe outputs are obtained from a non-allocableinput.11 One cannot disentangle the separatecontribution of each input to each product. Totalinput use is not determined by summing inputsused by each product as such a practice wouldresult in double counting.

Both characteristics of multifunctionalagriculture affect modeling farm-level and marketbehavior. If we assume competitive markets andignore the social value or cost of non-commodity

outputs, a farmer would maximize profits byequating the marginal value product of eachinput with its price. If it is possible to observethe levels of production of the non-commodityoutputs, then we also know that to maximizesocial welfare, the standard economicprescription is to ‘compensate’ producers by anamount equal to the marginal social value of apositive externality and ‘charge a fee’ equal tothe marginal social cost of a negativeexternality.

Although the logic of this Pigouvian principleis particularly compelling at a theoretical level, itspractical application is less straightforward. First,the levels of the externalities are difficult toobserve and measure. For this reason alone, itwould be difficult, if not impossible, to apply thePigouvian principle directly. Furthermore, incountries where markets are distorted bydomestic agricultural policy intervention to protectproducers’ income, or barriers to internationaltrade, Baumol and Oates (1988) were among thefirst to show that the Pigouvian policyinstruments would have to be set at levelsdifferent from their marginal social costs(benefits).12 To do a comprehensive evaluation ofthese multifunctional policies, it is important todetermine what levels of policy instrumentsensure the socially optimal supply of

(1B)

34

non-commodity outputs in the presence of thesedistortions.

The issue of non-observability of thenon-commodity externalities is addressed bytargeting multifunctional policy intervention atobservable variables that are correlated with theexternality generating process (e.g., Rude 2001;Romstad 2004), such as the inputs used foragricultural production. Indeed, in the absence ofmarket distortions, the first-best welfare scenariocan still be achieved if the appropriate fees(compensations) are applied to ‘all' inputscontributing to the production of thenon-commodity externalities or ‘public’ goods(Holtermann 1976; Griffin and Bromley 1982).13

Peterson et al. (2002) demonstrate this principlein the case where both positive and negativenon-commodity externalities are jointly producedwith agricultural commodity outputs.

In this report, we extend that type of analysisto the case where there are also market distortions.The appropriate modifications in the levels of theinput fees (compensations), for what is now asecond-best optimum due to the domestic orinternational policy intervention, are derived in amanner similar to that when the market distortion isin the form of imperfect competition.

Two-stage Approach to Policy Design

To derive the optimal second-best multifunctionalpolicy, we develop a two-stage approach arounda partial equilibrium model of an agriculturalmarket in which there are jointly producednon-commodity outputs. In the first stage, thegovernment implements the optimal policydesign by taxing or subsidizing land and other

inputs (Peterson et al. 2002).14 Given the levelsof the policy instruments announced by thegovernment, the representative rice-farmermakes an optimal decision on the use ofagricultural inputs in a second stage. Since thistwo-stage problem is solved through backwardinduction (Tirole 1988), it is convenient to beginwith a discussion of the second stage.

Optimal Decisions of Producers UnderDomestic Support Policy

To understand the implications of domesticsupport for optimal multifunctional policydesign, we must formulate the producer’s‘decision-problem’ within the context of thespecific design of agricultural support policy forrice in Taiwan. The strategy we use at present issimilar to what Fraser (2003) used in modelingthe effects of a reduction in domestic support foragriculture in Europe.

Taiwan has both a price support program(the limited purchase support program) and aland set-aside program (Huang 2001). There is apayment for each hectare set- aside, but thegovernment also limits purchases of rice at thesupport price to a fixed quantity per hectare.This is an important feature of the policybecause it is likely that the supply-inducing priceat the margin will be the domestic market-clearing price, rather than the support price. Thischaracteristic of the price support program mayreduce the level of the supply response as thesupport price is reduced.

The representative rice-farmer maximizesprofits; revenue includes market sales of rice,rice sold to the government at the support price,

13Shortle et al. (1998, pp. 574-575) also argue that these instruments provide the correct marginal incentives for input use, but do notguarantee that the set of firms left in the industry will be the efficient ones. Thus, they argue that an additional instrument, that doesnot distort input levels, may be needed to influence entry and exit of firms. This could be in the form of a lump sum tax on extra-marginal firms to ensure negative profits if they produce, or a compensation for not producing that would be larger than their after-taxprofits. Further, while no such tax would be required for marginal or infra-marginal firms, a lump sum compensation may be necessaryfor those firms whose entry or exit is influenced by the input taxes or subsidies. The necessity for this lump sum tax or compensationclearly has implications for the re-instrumentation of government policy where there remains a desire to support the incomes of agriculturalproducers.

14For simplicity, we consider only two inputs in this conceptual analysis. The results extend in a straightforward fashion to the fourinputs (fertilizer, labor, land and water) included in the empirical model of rice in Taiwan.

35

tZαLsZPLP

LPQPαLQαLZαLFPMax

zl

sZL

−+−−

−++−=

)(

)1(}),({,

απ

0)1(

)(

=−+−+

−+∂

∂=

∂

∂

ls PαsP

PPαQL

FαPL

π

α

−∂

∂=

∂

∂P

z – t = 0

Z

FP

Z

π

and payments for the set-aside land. The farmer’sdecision problem is:

(2B)

where F(.) is the production function for rice, P isthe equilibrium market price for rice, is themandatory proportion of total land, L, in riceproduction, (1-a) is the proportion of land enrolledin the set-aside program, Z is the other inputused in rice production, P is the per unit

government purchase price, Q is the government

purchase quantity of per hectare, Ps is the perhectare set-aside payment, P1 is the price of land,and P2 is the price of a purchased input.Assuming that land contributes to the productionof the positive externality, E1 from equation (1B),and Z contributes to the negative externality, E2

from equation (1B), s and t are the compensationand fee, respectively, on input use needed toensure the appropriate production of the non-commodity environmental output variables.15 Thefirst term in the profit function is market revenue;the second term is the revenue from governmentpurchases based on maximum per hectare

quantity Q; the third term is the revenue from the

land set-aside program, and the last two termsare the compensation revenue and fee cost to thefarmer associated with input use. Under the first-order, conditions necessary for an interior solutionare:

(3B)

(4B)

The economic intuition behind equations (3B)and (4B) may be demonstrated by comparing themarginal cost and benefit of the use of each input.

αL, Z]) – D(E2[αL, Z]) – αLQ(P - P)

SW = p(h) dh –p(Qc)Q + π (p, P, Q, Ps , t, s)

–

– –

–Qc

0

.

.

+B(E1 [

) LPs + tZ – s(αL)–(1– α

15The levels of these policy instruments are determined in the first stage of the ‘policy design problem’ described below.

For optimal land use, equation (3B), the first term isthe marginal revenue from selling rice on themarket; the second term is the marginal paymentreceived through the price support program; the thirdterm is the marginal payment for the land set aside;and the fourth term is the marginal compensationfor the contribution of land in production to thesupply of a ‘positive’ non-commodity environmentaloutput. The solution of the farmer’s optimizationproblem requires that these combined marginal (net)contributions of land to profit are equated to theprice of land. Assuming that the marginal product ofland is declining, and since the three termsinvolving policy instruments are positive, the optimaluse of land will tend to be higher than undercompetitive market conditions in the absence of aprice policy and an environmental policy thatexplicitly recognizes the social value of the positivenon-commodity externality. In contrast, we can seefrom equation (4B), that the optimal level of thepurchased input will tend to be below what it wouldbe in a competitive market because of theassumed effect of the use of the input on theoutput of the negative non-commodity externality.The value of the marginal product of Z is equatedto its price plus the fee on Z.

Optimal Levels of EnvironmentalPolicy Instruments

Given the solution to the rice-farmer’s profitmaximizing problem, second-best optimalenvironmental policies (given domestic policydistortions) can be derived by maximizing thesocial welfare function. Social welfare isrepresented by the sum of consumers’ andproducers’ surpluses, less governmental budgetcosts and the net value of the non-commodityenvironmental externalities. Net social welfare canbe expressed as:

α

(5B)

36

)}

} = 0

)1(

)][({ 21

sP

QPPsL

ED

L

EB

s

L

s

SW

−−

−−−

t+

∂

∂′−

∂

∂′

∂

∂=

∂

∂ α

{

2

1

Z

ED

Z

EB

s

Z

∂

∂′−

∂

∂′

∂

∂+α

}

} = 0

)1()[

({ 21

sPQP]P

sL

ED

L

EB

t

L

t

SW

−−−−

−

t+

∂

∂′−

∂

∂′

∂

∂=

∂

∂ α

{ 21

Z

ED

Z

EB

t

Z

∂

∂′−

∂

∂′

∂

∂+

α

= 0)1()[

21

sPQP]P

sL

ED

L

E(B

−−−−

−∂

∂′−

∂

∂′α

α

t12

Z

EB

Z

ED

∂

∂′−

∂

∂′=

where p (h) is the demand curve for rice, P (Qc)

is the equilibrium price, and � is the farmer’sindirect profit function. B (.) is the total benefitfunction associated with the positiveenvironmental externality and D (.) is the totaldamage function associated with the negativeexternality.16 The first two terms of equation (5B)represent consumers’ surplus; the third term isthe profit of the representative rice-farmer, thefourth and fifth terms are the benefit and damagefunctions associated with the positive andnegative non-commodity externalities. Thesenon-commodity outputs are supplied jointly with agiven level of inputs in rice production. Theremaining two terms represent government feesand compensations, respectively, on the twoagricultural inputs.

We apply Shephard’s lemma to determine theoptimal levels of the two new policy instrumentsby partially differentiating equation (5B) withrespect to t and s. The first-order necessaryconditions for an interior solution to themaximization of social welfare are:

Making the reasonable assumption that thepartial derivatives of the input levels with respectto the fee and the compensation are nonzero, thelevels of t and s that satisfy these first-orderconditions also satisfy:

(8B)

.(9B)

Solving equations (8B) and (9B) for t and s,we have:

(10Ba)

(10Bb)

The intuition behind these two equations(10Ba and 10Bb) is straightforward; the optimalinput compensation (fee) is determined in part bythe products of the marginal contributions of eachinput to the production of the externalities and themarginal benefit or damage of each externality.Since both inputs contribute to both externalities,it is impossible to determine 'ex ante' if thecompensation or fee is positive or negative. Forexample, the optimal fee for the non-land inputmay well be negative if that input’s marginalcontribution to the benefits associated with thepositive externality outweighs its marginal

16Two issues must be addressed in specifying these environmental benefit and cost functions. The first has to do with the difficultiesassociated with the valuation of jointly produced nonmarket goods that are discussed in detail by Boisvert et al. (2004). Although,Randall (2002) has proposed a strategy for addressing these issues, no empirical work is available. Therefore, in our empirical analysis,it is necessary to assume that the cross derivative between E1 and E2 in the social value function for the externalities (V [E1, E2, I],where I is real income), is small. Further, our policies are likely to have only a small effect on the national income. Thus, we approximate

the value function as V � B (E1) – D (E2). The second issue relates to the Es being unobservable. If we assume that they arestochastic and contingent on both L and Z, B (.) and D (.) can be replaced by their expected values; appropriate fees and compensationsapplied to all inputs affecting the Es are preferred to taxes and subsidies applied to forecasts of the Es (Shortle and Dunn 1986).

(6B)

(7B)

and

and

and

= 0t+21

Z

ED

Z

EB

∂

∂′−

∂

∂′

s 21

L

ED

L

EB

∂

∂′−

∂

∂′= –

1

)( QPP

−−

−

αα Ps

37

contribution to the cost of damage associatedwith the negative externality. Whether thecompensation on land is positive or negativedepends on similar considerations, but, inaddition, the amount of the land compensationdepends in part on the distorting effect of thelimited price support and the land set-asidepayment. To underscore the effect of marketdistortions on the level of ‘Pigouvian-type’ policyinstruments, the land compensation that ensuresthe optimal level of the multiple externalities willdecrease if the level of domestic support (eitherthe price support or the set-aside payment)increases (equation [10Bb]). Without domesticsupport, the optimal compensation for land isequal to the net effect of land’s net marginalcontribution to both externalities.

Opening the Economy to InternationalTrade

The optimal environmental policy designrepresented in equations (10Ba and 10Bb) appliesto a closed economy, and was applicable to therice market prior to Taiwan’s admission to theWTO, which, in 2003, led to rice imports under atariff-rate-quota (TRQ). For this reason, we mustalso determine how the optimal multifunctional

fees and compensations are affected by the newpolicy regime.

To understand the TRQ, we follow the argumentdeveloped both algebraically and graphically byAbbott and Paarlberg (1998) for a small importingcountry, (e.g., the situation for rice in Taiwan). Tomodel the TRQ regime, we must recognize thatthere are three possible outcomes. Under the first,where imports equal the minimum accesscommitment, the TRQ acts like a quota in which atariff is also levied. Accordingly, the optimalenvironmental policy for this small open economy issimilar to that in equations (10Ba and 10Bb), exceptthat the domestic price is now determined by thesum of domestic production plus imports, ratherthan by just domestic production as in the closedeconomy case. Second, if desired imports areless than the minimum access commitment, thenthe below quota tariff would be effective. TheTRQ acts like a pure tariff in this case, and thedomestic price is the world price plus the withinin-quota tariff. For the third outcome, imports canexceed the minimum access level—in which casethe higher out-of-quota tariff would apply.17 As isseen in the empirical analysis above, it is thefirst outcome, where imports equal the minimumaccess commitment and the TRQ acts like aquota in which a tariff is also levied, that isapplicable to the Taiwanese rice.

17Both Krutilla (1991) and Peterson et al. (2002) demonstrate that effects of these types of trade policies are slightly different for thelarge country case; the government can exploit the terms-of-trade effect of policy intervention.

Research Reports

1. Integrated Land and Water Management for Food and Environmental Security.F.W.T. Penning de Vries, H. Acquay, D. Molden, S.J. Scherr, C. Valentin andO. Cofie. 2003.

2. Taking into Account Environmental Water Requirements in Global-scale WaterResources Assessments. Vladimir Smakhtin, Carmen Revenga and Petra Döll.2004.

3. Water Management in the Yellow River Basin: Background, Current Critical Issuesand Future Research Needs. Mark Giordano, Zhongping Zhu, Ximing Cai, ShangqiHong, Xuecheng Zhang and Yunpeng Xue. 2004.

4. Does International Cereal Trade Save Water? The Impact of Virtual Water Tradeon Global Water Use. Charlotte de Fraiture, David Molden, Mark Rosegrant,Ximing Cai and Upali Amarasinghe. 2004.

5. Evolution of Irrigation in South and Southeast Asia. Randolph Barker and FrançoisMolle. 2004.

6. Macro Policies and Investment Priorities for Irrigated Agriculture in Vietnam.Randolph Barker, Claudia Ringler, Nguyen Minh Tien and Mark Rosegrant. 2004.

7. Impacts of Irrigation on Inland Fisheries: Appraisals in Laos and Sri Lanka. SophieNguyen-Khoa, Laurence Smith and Kai Lorenzen. 2005.

8. Meta-Analysis to Assess Impact of Watershed Program and People’s Participation.P.K. Joshi, A.K. Jha, S.P. Wani, Laxmi Joshi and R.L. Shiyani. 2005.

9. Historical Transformations of the Lower Jordan River Basin (in Jordan): Changesin Water Use and Projections (1950–2025). Rémy Courcier, Jean-Philippe Venotand François Molle. 2005.

10. Cities versus Agriculture: Revisiting Intersectoral Water Transfers, Potential Gainsand Conflicts. François Molle and Jeremy Berkoff. 2006.

11. Prospects for Productive Use of Saline Water in West Asia and North Africa.John Stenhouse and Jacob W. Kijne. 2006.

12. Impact of Land Use on River Basin Water Balance: A Case Study of the ModderRiver Basin, South Africa. Y.E. Woyessa, E. Pretorius, P.S. van Heerden, M.Hensley and L.D. van Rensburg. 2006.

13. Intensifying Agricultural Sustainability: An Analysis of Impacts and Drivers in theDevelopment of ‘Bright Spots’. A. D. Noble, D. A. Bossio, F. W. T. Penning deVries, J. Pretty and T. M. Thiyagarajan. 2006.

14. Multifunctional Agricultural Policy, Reduced Domestic Support and LiberalizedTrade: An Empirical Assessment for Taiwanese Rice. Richard N. Boisvert andHung-Hao Chang. 2006.

ISSN 1391-9407ISBN 92-9090-648-0ISBN 978-92-9090-648-3

Postal Address: IWMI, P. O. Box 2075, Colombo, Sri Lanka Location: 127, Sunil Mawatha, Pelawatte, Battaramulla, Sri LankaTelephone: +94-11 2787404, 2784080 Fax: +94-11 2786854

Email: comp.assessment@cgiar.org Website: www.iwmi.org/assessment