David Orden, Clare Narrod and Joseph W. Glauber

October 2000

Draft paper presented at the workshop on "The Economics of Quarantine" sponsored by Agriculture, Fisheries and Forestry-Australia, Melbourne, Australia, October 2000. David Orden is professor of agricultural and applied economics at Virginia Polytechnic Institute and State University, Blacksburg, VA 24061, USA (orden@vt.edu). Clare Narrod is AAAS Risk Policy Fellow, Office of Risk Assessment and Cost-Benefit Analysis, U.S. Department of Agriculture, Washington D.C. 20250, USA (cnarrod@mailoce.oce.usda.gov). Joseph Glauber is Deputy Chief Economist, U.S. Department of Agriculture, Washington D.C. 20250, USA (joseph.glauber@usda.gov).

This paper examines the choice among sanitary and phytosanitary policies of those that are least trade distorting. In addressing this choice, we highlight the potential for complementarity between science-based risk assessment and economic-based benefit-cost analysis in regulatory decision processes. We make the argument for fuller integration of these approaches than is often the case. Integrating risk assessment and benefit-cost analysis simultaneously into the regulatory process provides decision makers with a rich two-dimensional nexus of information. It is too optimistic to expect that for all regulatory decisions a fully optimal policy choice can be achieved when only a single dimension of information is considered. The two-dimensional risk assessment-economic analysis nexus gives decision makers an opportunity to evaluate the trade-offs that are faced when they choose among alternative regulatory measures. The criterion "least trade restrictive" (or more generally, "least trade distorting") is one that policy makers can apply to these decisions. It is not a complete decision making rule, nor is it the only criterion on which policy options might be ranked, but least trade restrictive is a criterion mandated by the WTO for consideration in SPS policy determination.

The paper is organized as follows. The next section provides a brief discussion of the concept of a policy being least trade distorting. We follow by summarizing results from two case studies that were reanalyzed using somewhat different approaches than those utilized in regulatory decisions. The possibility for either convergence or divergence between the inferences drawn from risk assessment versus benefit-cost analysis is demonstrated for the case of regulation of avocados entering the United States from Mexico. The value of integrating benefit-cost analysis and risk assessment simultaneously into evaluation of risk management options is demonstrated for the case of setting internal U.S. Karnal bunt quarantine rules. We draw on these two case studies in a final section to offer some conclusions about the criterion "least trade distorting" and about integrating risk assessment and benefit-cost analysis into regulatory decision making. The conceptual framework for identifying a policy that is economically efficient as being least trade distorting is well defined. The hard part is providing the risk and economic assessments on which such decisions rest in specific cases. In addition, decision makers may not want to consider the full range of options that risk assessment and benefit-cost analysis lay before them.

What does it mean to say that a policy is least trade distorting, and why is this phrase so appealing to trade negotiators, but less so to trade economists? The answers to these questions come from recognizing that being least trade distorting is only one attribute of a policy being most efficient at achieving a stated objective. That there is an objective other than trade liberalization which the policy is intended to achieve is implicit in the notion that a given policy is least trade distorting. But achieving that policy objective may itself not be optimal on economic grounds, such as those often used to argue for benefits of liberalized trade.

The general theory of economically efficient policy intervention suggests that policies be directed specifically at the given objective, because such policies attain their ends without imposing other distortions on the economy. A subsidy to producers is a more efficient policy for increasing output than a tariff, for example, since the production subsidy does not raise prices paid by consumers, and thus avoids distorting their optimal consumption decisions along with production levels. Likewise, if the policy objective is to subsidize use of labor, a wage subsidy is more efficient than a general production subsidy, which is more efficient than a tariff. In short, intervention policies should be targeted toward the intended objective, and the more precisely they can be targeted, the better. The more efficient the policy intervention, the less trade distorting that policy will be for any given level of the objective achieved by the intervention.

The theory of efficient policy intervention applies equally well to interventions aimed at correcting market failures. For example, if domestic production causes a negative local externality, a tax on production is the efficient policy, as opposed, for example, to an import subsidy that lowers the product price to producers and consumers. If the use of a particular input causes a negative externality, then the optimal policy instrument would be a tax on use of that input, not a tax on production per se.

There is another consideration when policies are directed at externalities. When there is no externality, there is no optimal objective of the policy intervention inherent in standard economic welfare analysis; instead the level of the intervention has to be presumed to be set exogenously, or to emerge from an implicit political economy model. When there is an externality, not only is there a corresponding optimal policy instrument, but there is an inherently optimal level of intervention that maximizes economic welfare. Policy can now err by applying the wrong instrument, but also by applying the optimal instrument at too low or too high a level.

Sanitary and phytosanitary regulations on trade are optimal policy instruments when trade of a commodity is associated with the risk of incurring a negative externality of a deleterious effect on domestic plant, animal or human health. The SPS agreement makes it perfectly clear that SPS regulations are not acceptable policies for achieving other objectives. They would not be least trade distorting policies for these other objectives, and their use would raise additional problems.

Suppose one approaches the issue of SPS regulations along only the risk assessment dimension. If the mandate of regulatory authorities to protect the domestic economy from negative SPS externalities is stated in strong terms, as it often is, then product bans and other severe quarantine measures emerge quite naturally as policy outcomes. A product ban is a high level of intervention to address an SPS externality, but a ban does eliminate the externality risk to the extent that trade is its proximate cause. Within the risk assessment dimension, there is room for dispute over whether an externality threat exists in a given situation. And a ban may or may not be least trade distorting—perhaps there is another way to eliminate the externality risk, one that allows the product to be traded under some specified conditions. Either way, when the policy decision is perceived only in the risk assessment dimension, there is no impetus to ask whether the cost of the policy is warranted by the expected benefits; that is, whether the level of intervention needed to achieve the risk-reduction objective is also desirable on economic criteria, such as maximizing the expected contribution of the affected markets to national welfare.

There are dramatic alternatives to SPS-risk-induced policies such as product bans. If strong property rights could be assured to those who might suffer the damages from an externality, and if insurance markets were sufficiently developed, then one could do away with many SPS regulations, and let market outcomes evolve unfettered, with the externality internalized by the assignment of property rights. Neither WTO laws, nor other international laws, nor domestic liability laws, nor risk-sharing markets are strong enough for this option to emerge in the short term.

Instead, the challenges to existing SPS regulations are coming primarily as requests for easing of the most severe trade-restricting policies. The key to these alternatives is often a systems approach to risk management, whereby a set of procedures are specified that in principle reduce the externality risk associated with trade of a commodity. Requests for adoption of systems approaches rest on a firm foundation in the WTO Agreement on the Application of Sanitary and Phytosanitary Measures (WTO 1994). Specifically, Article 5.6 states that:

"Without prejudice to paragraph 2 of Article 3, when establishing or maintaining sanitary or phytosanitary measures to achieve the appropriate level of sanitary or phytosanitary protection, Members shall ensure that such measures are not more trade-restrictive than required to achieve their appropriate level of sanitary or phytosanitary protection, taking into account technical and economic feasibility.³"

The footnote 3 in the Agreement states that:

"For purposes of paragraph 6 of Article 5, a measure is not more trade-restrictive than required unless there is another measure, reasonably available taking into account technical and economic feasibility, that achieves the appropriate level of sanitary or phytosanitary protection and is significantly less restrictive to trade."

Three basic questions about policy decisions arise when a systems approach is considered. A set of procedures to be applied to a commodity in order for it to qualify for importation are specified with the objective to reduce risk, perhaps to (essentially) zero, equivalent to a product ban. First, a risk assessment question arises: does the specified set of procedures achieve the risk objective? Second, a least trade distorting question arises: does the specified set of procedures distort trade the least among possible options? Finally, a larger economic efficiency question arises: is the targeted level of risk reduction itself justified by the economic welfare effects that are associated with it? The latter two questions bring the benefit-cost dimension into policy considerations, along with the risk assessment dimension.

Our first case study illustrates the potential for convergence or divergence between policy decision criteria based on risk assessment versus benefit-cost analysis. In many cases, sound science and sound economics will correspond—as when pest risks have expected economic costs that outweigh the benefits of trade. These cases provide an implicit economic rationale for the scientific focus of international SPS agreements. However, some SPS regulations may have economic net costs even if they have solid scientific justification. In these latter cases, a regulation is not good economic policy in terms such as maximizing expected national welfare, even if the regulation does not violate the SPS agreement of the WTO.

One of the most contentious SPS disputes between Mexico and the United States concerns U.S. restrictions on importation of Hass avocados. Mexico argues that its principal avocado producing region has low incidence of pests of quarantine significance, that the Hass avocado is not a preferred host for some pests of concern, and that a systems approach to handling fruit for export has proven effective in eliminating risks of pest infestations being carried abroad. The U.S. avocado industry, which is concentrated in southern California, bitterly contests opening its domestic market to exports from Mexico. The industry acknowledges that it receives prices well above those of Mexican exports, but argues that its fear is not competition in the marketplace but risks of pest infestations associated with trade. Domestic U.S. producers challenge Mexican assessments of pest risks and the effectiveness of the systems approach to risk management.

The U.S. Department of Agriculture is caught in the middle of this controversy. Its Animal and Plant Health Inspection Service (APHIS) and Agricultural Research Service (ARS) have engaged in intense bi-national technical negotiations with Mexican authorities about pest risk evidence and export protocols that might sustain easing of the import ban. Following four years of negotiations, in September 1994, APHIS accepted a Mexican work plan proposing a systems approach to pest risk mitigation. With some further safeguards, a proposed rule was published by USDA in July 1995 to allow imports into the northeastern United States of Mexican avocados grown and processed under specified conditions. Imports were to be limited to the winter months when the risk of establishment of pests is further mitigated by adverse weather.

The geographic and seasonal restrictions of USDA’s proposed rule implied that the partial easing of the avocado import ban opened less than five percent of the annual U.S. market to Mexican products. The domestic avocado industry fought against even this limited trade rule, but in February 1997 USDA announced it would ease the avocado ban. Since that time, limited avocado trade has been allowed. By 2000, Mexico requested additional access to the U.S. market, and subsequent regulatory assessments are pending.

To evaluate the economic impacts of U.S.-Mexico avocado trade, Orden and Romano (1996) examined the effects of full or partial easing of the import ban. Estimates are derived of a linear supply function that is inelastic in the short run (0.28, when lagged quantity is held constant) and elastic in the long run (1.18, when quantity is in a steady state). A linear estimate of demand is inelastic (-0.45) but estimation of a nonlinear demand specification yields a price flexibility of -0.65, corresponding to an elasticity of -1.53. Thus, the estimated supply and demand functions provide point estimates that span a range from inelastic to elastic behavioral responses. Orden and Romano applied the elastic values as a long-run model and the inelastic values as a short-run model in which producers and consumers are less price responsive. For both models, the assumption is made that Mexican supply is perfectly elastic at the wholesale price for delivery of avocados from Mexico to New York of $878/ton, as calculated by Garoyan (1995). The assumption of a perfectly elastic supply is most plausible for a partial easing of the import ban. It is arguably an oversimplification for evaluation of the effects of the quarantine being removed completely, since the expanded traded might put upward pressure on the Mexican price.

In Orden and Romano’s models of partial easing of the avocado import ban, the domestic U.S. market is divided into two submarkets—the northeastern regional winter market and the national aggregate for all other regions and seasons. The quantity of California avocados shipped to the northeastern region during the four winter months at prevailing domestic prices was reported by Garoyan to have averaged 3,819 tons during 1986-95. In modeling limited trade, the domestic price in the northeastern winter regional market is assumed to fall to the free-trade level for imports from Mexico, inducing greater consumption than at higher past domestic prices. An aggregate price for the rest of the U.S. market is determined by an equilibrium of domestic supply and demand with the northeastern winter demand excluded.

Estimates of the probabilities of pest infestations are pivotal to regulatory decisions about avocado imports. Firko (1995) made the estimates utilized by APHIS. Among four potential pests (fruit flies, seed weevil, stem weevil, and seed moth), he estimated that the maximum probability of an infestation occurring in the United States for partial easing of the import ban under a systems approach to risk mitigation was p_AM = 0.00345, the probability of a pest infestation associated with the introduction of stem weevil. Firko estimated that the probability of infestation of stem weevil had a minimum value p_Am = 1.35x10^-6. These risk estimates were considered too low by the domestic industry. Nyrop (1995) calculated that the time expected to pass before an infestation of stem weevils occurred under the 1995 proposed rule ranged from less than one year to 20 years. The corresponding probabilities of pest infestation were treated by Orden and Romano as p_NM = 1.0 and p_Nm = 0.05. The four alternative probability estimates from Firko and Nyrop (AM, Am, NM, Nm) were used to characterize the range of risks of pest infestation (from essentially zero to certainty) that might be associated with either partial easing or complete removal of the ban.

The final parameters affecting the economic analysis are estimates of the costs from increased production expenses and lost productivity associated with a pest infestation, which are modeled as a proportional shift in the domestic supply function. Evangelou et al. (1993) had estimated that weevil infestation would cause a 41 percent increase in marginal cost due to increased application of pesticides and a 20 percent reduction in yield, but argued these estimates were somewhat overstated. Thus, Orden and Romano considered several possible impacts on production. The largest impacts were assumed to involve a 60 percent increase in marginal costs and a 20 percent reduction in yield (denoted 60-20). The smallest impacts were assumed to be a 20 percent increase in marginal costs with no reduction in yield (20-0).

The effects of free trade when a pest infestation might adversely affect domestic production are illustrated in figure 1. The domestic price P_D1 falls to the world price P_W and consumer surplus increases (by C+D+E) whether or not an infestation occurs. Producer surplus falls by C+D (the trade effect) and additionally by G (the infestation effect) if pests raises production costs and lower yields with certainty, shifting domestic supply from S to S'. Consumers are always better off, producers are always worse off, and the net effect on welfare (E-G) can be positive or negative. On a probabilistic basis, the expected domestic supply function will lie between S and S', with its location depending on the assumed level of pest infestation risk.

The analysis is more complicated when only limited quantities of imports are allowed, as shown in figure 2. Ignoring regional submarket considerations, which are not depicted in figure 2 for simplicity, the limited imports would lower the domestic price from P_D1 to P_D2 if there were no pest infestation, rather than to the world price level. Pest infestation reduces domestic supply and affects the domestic price in the opposite direction from imports. With limited trade, the equilibrium domestic price can rise or fall. When the domestic price rises, as shown from P_D1 to P_D3 in figure 2, consumers are worse off (by c+d). Producers surplus rises (by c) with the higher price but falls due to higher production costs (by f+i+k). Producers may be better or worse off than at the initial equilibrium (better if c>f+i+k). Producers may also be better or worse off than with trade but without a pest infestation (better if c+e>i+k). Whatever the outcome for producers, social welfare falls (by d+f+i+k) compared to its level at the initial equilibrium, or compared to its level at price P_D2 with trade but without pest infestation (by d+f+i+k+g).

The expected economic effects of trade are shown for the avocado case in table 1 for a long-run model with estimated elastic supply and (nonlinear) demand. The initial equilibrium with avocado imports prohibited occurs in this estimated model at a domestic price of $1385 and output of 132,430 tons. Consumer surplus is $134.4 million and producer surplus is $91.6 million. If trade were completely liberalized and no pest infestation occurs, the domestic price falls to $878, consumption increases to 222,722 tons, and domestic production declines to 83,904 tons. Consumer surplus rises by $87.5 million, producer surplus falls by $55.2 million, and the net welfare gain is $32.4 million (14 percent of initial consumer plus producer surplus).

A pest infestation exacerbates the adverse effects of free trade on domestic producers through a lower price, and reduces the net welfare gain. In the worst case scenario of certain infestation (p_NM = 1.0) and highest costs (60-20), producer surplus falls by an additional $18.4 million in the long-run model. There remains a net welfare gain even in this case, although it is reduced to $13.9 million. Thus, even when free trade is bad phytosanitary policy, it is good economic policy, in the sense of raising net national welfare. For probabilities of pest infestation at Nyrop’s minimum (Nm) or lower, the effect of an infestation on expected producer surplus is less than $2 million, and the expected net welfare gain remains above $30 million.

The partial easing of the avocado import ban under USDA’s rule has smaller economic effects than free trade when no pest infestation occurs, as shown in the lower half of table l. The net national welfare gains is $2.5 million (about 2 percent of initial total consumer plus producer surplus). In the northeastern region, winter consumption increases and consumer surplus rises by $2.5 million (not shown separately in the table) as the price falls to that of exports from Mexico. The domestic price for the aggregate U.S. market with the northeastern winter demand excluded falls by 1.3 percent (from $1385 to $1368), as domestic consumption displaced from the northeastern winter market is absorbed by a combination of expanded consumption elsewhere and reduced domestic supply. Consumer surplus increases by $2.2 million outside of the northeast, but producer surplus falls by a similar amount (the net welfare gain is only $33,337 outside of the northeastern winter market). Thus, the limited opening of trade under the proposed partial easing of the import ban has positive effects on northeastern winter consumer surplus, and limited positive effects on other consumers and negative effects on domestic producers.

When imports are restricted under partial easing of the import ban, increased marginal costs and lowered yields reduce producer surplus by $45.8 million for the worst-case scenario of a pest infestation. The reduced supply pushes the equilibrium domestic price up (excluding the northeastern winter market) from $1385 to $1795. The price increase offsets $31.1 million of the loss of producer surplus, leaving a net loss of $14.7 million, still almost seven times as large as the effect from limited trade alone. A larger economic effect of the pest infestation is felt by consumers outside of the northeastern winter market. With the increased domestic price in the worst-case scenario, their consumer surplus falls by $43.5 million. Thus, negative economic impacts of pest risk are borne by consumers outside the northeastern winter market as well as producers when trade is opened only to a limited extent.

The potential losses to consumers and producers under certainty of a pest infestation are large enough that the net welfare loss is still $13.6 million under the lowest assumed costs to production (20-0) from the infestation. Thus, for high probabilities of pest risk, the limited easing of the avocado import ban is both bad phytosanitary policy and bad economic policy. Under the assumption of highest pest-infestation costs, expected consumer surplus rises at risk probabilities as high as the Nyrop minimum risk level (Nm), but the expected gains of consumer surplus is less than the expected loss of producer surplus with limited trade for this level of pest infestation risk. Only at lower risk levels, or lower infestation costs, is a partial easing of the import ban a regulatory decision that raises expected net economic welfare.

Sensitivity of the economic results to the elasticity assumptions is illustrated by comparing the above results to the outcomes from a short-run model with inelastic estimates of supply and demand, shown in table 2. For the estimated short-run model, the initial equilibrium with avocado imports prohibited occurs at a domestic price of $1950 and output of 140,496 tons. Consumer surplus is $189.1 million and producer surplus is $230.9 million. Under the assumption of free trade without pest infestation, consumer surplus increases by $180.5 million, producer surplus falls by $137.6 million, and the net welfare gain is $43.0 million (10 percent of the initial sum of consumer and producer surplus). Pest infestations compound losses of producer surplus under free trade in the short-run model, but again there is a net welfare gain even when pest infestation occurs with certainty and has high cost.

As before, the effects on producers and consumers with limited trade and no pest infestation are much smaller than under free trade. The domestic price (outside the northeastern winter market) falls to $1899 in the short-run model, total consumer surplus increases by $12.2 million, producers surplus falls by $7.1 million, and the net welfare gain is $5.1 million. For the worst-case scenario of certain pest infestation and high costs, the domestic price is pushed up to $2540 in the short-run model with limited imports. Consumer surplus falls by $66.6 million (gain of $5.2 million in the northeastern winter market, but loss of $71.8 million elsewhere). Partial easing of the import ban is only good economic policy when low risk probabilities and costs are assumed, as in the long-run model.

With the domestic price pushed up to $2540 in the short-run worst-case scenario, producers are better off when limited imports are associated with a pest infestation than when limited trade occurs without an infestation (producer surplus is greater by $8.5 million). Generally, when trade is only partially opened, producers are expected to be better off in the short-run model the higher the probability of pest infestation (this was not the case in the long-run model). Producers are even slightly better off with limited trade and pest infestation than they are at the initial equilibrium under the assumption of highest costs of an infestation.

To summarize, the economic analysis for U.S. policy on imports of Mexican avocados suggests that free trade would raise consumer surplus, lower producer surplus, and increase national welfare even if pest infestations were certain to occur. With the partial easing of the ban and limited imports that have been adopted, expected consumer and net welfare gains from trade are relatively small and can be exceeded by the expected costs of pest infestation when risks of infestation are high. With limited trade and high probabilities of pest risk, consumers bear more of the economic costs from the risk of pest infestation than do producers. At lower pest-risk levels, expected consumer surplus increases, and the expected gains offset expected producer surplus losses, so expected net welfare rises. In the long-run model (with elastic supply and demand) pest infestations add to the losses of producers that result from allowing limited trade. But in the short-run model (with inelastic supply and demand) the increased producer surplus from higher domestic prices more than offsets the loss of producer surplus from higher costs and lower yields, so pest infestation risks have the net effect of lessening the decrease in expected producer surplus compared to limited trade without pest infestation.

The preceding analysis illustrates the possibilities for convergence or divergence between risk-reduction objectives versus benefit-cost objectives, but it does not explicitly bring benefit-cost analysis to bear on the design of quarantine procedures. Doing so requires an evaluation of how various aspects of the systems approach affect risk reduction, and the expected levels of benefits and costs realized. For avocados, limiting imports on a geographic and seasonal basis were one component of the systems approach to risk mitigation. An approximation to the upper bound on the efficacy of this dimension of risk management can be derived under the assumption that an infestation is certain to occur with full trade liberalization, whereas with limited trade the risk is reduced to the minimum level determined by Firko. Under this crude assumption, the trade limitations are effective as part of the systems approach to risk reduction, but achieving that risk reduction reduces expected welfare gains.

Glauber and Narrod (2000) present a more systematic analysis of the marginal risk reduction effects and expected benefits and costs associated with various protocols of an SPS regulation. They re-work the original USDA risk analysis used to design a quarantine to prevent the spread of Karnal bunt within the United States in 1996. This case study illustrates how benefit-cost analysis can be integrated with risk assessment to assess the marginal efficacy of specific components of a quarantine policy.

Karnal bunt is a fungal disease affecting wheat, rye and triticale. While posing no risk to human health, Karnal bunt can cause production losses to wheat in the form of reduced yields due to the infestation of kernels and reduction in the quality of wheat flour. Karnal bunt was first detected in the United States (in Arizona, New Mexico and Texas) in March 996. As a result, USDA quickly established testing for Karnal bunt teliospores, and regulations that quarantined all of Arizona and portions of New Mexico and Texas to avoid spread of the disease to other wheat growing regions of the country. Glauber and Narrod provide a detailed account of the regulation associated with the quarantine. They also review the risk assessments by Podleckis and Firko (1996a,b,c,d) that were used in design of the quarantine, and the benefit-cost analysis that was conducted.

In their analysis, Glauber and Narrod recreate the original Karnal bunt risk model, which focused on measuring risk of individual potential pathways for disease spread. With movement of positive-tested grain and seed outside of the quarantine area prohibited, these pathway for spread of the disease from negative-tested wheat still included: 1) millfeed (by-products of wheat milling that is fed to cattle); 2) transporting grain from the quarantined area to domestic grain storage facilities, mills and export elevators; 3) use outside the quarantine area of combines and other harvesting machinery; 4) infection through railcars that had transported infected wheat; and 5) planting of wheat seed originating in the quarantined area. These pathways led to regulations being considered on the following articles: 1) farm machinery and equipment used to produce wheat; 2) conveyances from field to handler, such as farm trucks and wagons; 3) grain elevators, equipment and structures at facilities that store and handle grain; 4) conveyances from handler to other marketing channels, such as railroad cars; 5) plant and plant parts, such as grain for milling, grain for seed, and straw; 6) flour and milling byproducts; 7) manure from animals fed wheat/wheat byproducts from the quarantine area; 8) used sacks; 9) seed-conditioning equipment; 10) byproducts of seed cleaning; 11) soil-moving equipment; 12) root crops with soil; and 13) soil.

Glauber and Narrod modified the initial risk assessment analysis to examine the overall level of risk of a Karnal bunt outbreak from any source originating in the quarantined area. They estimated the probability of at least one outbreak of Karnal bunt occurring outside the quarantined area, p*, as:

p* = 1 - (1 - p₁)(1 - p₂)(1 - p₃)(1 - p₄)(1-p₅)

where:

p₁ = probability of an outbreak of Karnal bunt outside the quarantined area from millfeed;

p₂ = probability of an outbreak of Karnal bunt in host fields outside the quarantined area

from grain in transit to mills or export elevators;

p₃ = probability of an outbreak of Karnal bunt outside the quarantined area from

combines or other harvesting machinery;

p₄ = probability of an outbreak of Karnal bunt outside the quarantined area from railcars

after grain is unloaded at mills or export elevators;

p₅ = probability of an outbreak of Karnal bunt outside the quarantined area from seed.

Glauber and Narrod found that the probability of outbreak via a given pathway was positively correlated with the number of railcars or other conveyances transporting grain or seed outside of the quarantined areas. Thus, a higher infestation of Karnal bunt within the quarantined area would mean less negative-tested wheat available for export or domestic milling purposes, lowering the probability of outbreak outside of the quarantined area. This interaction was incorporated into the risk assessment.

Glauber and Narrod’s analysis addresses the sensitivity of the overall level of risk of a Karnal bunt outbreak outside of the quarantine area, and finds it to be mostly influenced by the riskier pathways. Changes in the probability of outbreak in a given pathway with relatively low probability may be large in absolute terms, but have little effect on the overall level of risk. By focusing on individual pathways, the risk reducing potential of a specific protocol of the quarantine may be overestimated. For example, in the initial analysis a controversial requirement to heat-treat millfeed to eliminate risk of spread of Karnal bunt through manure of animals consuming this by-product was justified by USDA on the basis of the relatively sharp reduction in the risk of outbreak from contaminated millfeed from the heat-treatment procedure. When this factor is isolated, the results indicate that the millfeed treatment requirement reduced the mean risk of Karnal bunt outbreak from contaminated millfeed from 1 in 15,674 to 1 in 68 million. Yet the effect of the heat-treatment protocol was negligible in reducing the overall level of risk. Likewise, restrictions on the movement of negative-tested seed outside the quarantine area had a relatively small effect on the overall risk of outbreak.

In their reassessment of the original risk analysis for the Karnal bunt quarantine, Glauber and Narrod consider eight quarantine options by which risk of spread of the disease could be reduced. These options are based on four basic protocols. The first protocol restricted the movement of positive-tested grain and seed outside the quarantine area, but allowed all negative-tested grain and seed to move without significant additional restrictions. The second protocol required that all railcars be cleaned after delivery of wheat from the quarantined area. The third protocol restricted the movement of negative-tested seed outside of the quarantine area. The fourth protocol required the heat treatment of millfeed from quarantine-area wheat. These protocols were chosen as the focus of the analysis because they imposed the largest costs on the wheat industry in the Southwest when the quarantine was imposed and, as a result, were controversial.

Stochastic assessments of the effects of the eight options on reducing the risk of an outbreak of Karnal bunt outside the quarantine area are shown in table 3. The options are based on the four protocols, singly and in combination. The baseline (option 1) reflects the least restrictive policy, with the quarantine protocol limited to restrictions on the movement of positive-tested grain and seed. Grain and seed that twice tested negative for Karnal bunt teliospores would be free to move to domestic and export locations with no additional restrictions. Railcars would not be required to be cleaned. Options 2, 3 and 4 consider the other protocols individually in combination with the baseline restrictions. Of the individual protocols considered, railcar cleaning (option 2) had the largest effect on the overall level of risk of outbreak because of the relatively high risk of contamination through railcars. Restrictions on the movement of negative-tested seed (option 3) and millfeed treatment requirements (option 4) had minimal effects on the overall level of risk. Options 5 through 8 include combinations of the proposed protocols. Among these, option 8 reflects the system put in place by APHIS following the discovery of Karnal bunt in Arizona. Taken together, the protocols in option 8 reduced the mean level of risk by 97 percent compared to the baseline (option 1).

The original USDA regulatory impact analysis assumed that failure to implement the quarantine would jeopardize U.S. exports to those countries that maintained restrictions against wheat from Karnal bunt infected countries at the time the disease was discovered in Arizona. The United States was exporting about 1.2 billion bushels of wheat annually in 1995-96, with an estimated value of $3 to $4 billion. About one-half of U.S. wheat exports were shipped to countries that had Karnal bunt restrictions. Thus, effects on the wheat market from spread of the disease beyond the quarantine area were considered that ranged from a 10-percent net loss of export markets to a 50-percent loss. For example, a decrease of 10 percent in exports was estimated to cause a $0.22 per bushel drop in the wheat price and reduce wheat sector income by over $500 million per year. A decrease of exports of 50 percent was estimated to cause the price of U.S. wheat to fall by 30 percent and lower net sectoral income by $2.7 billion annually. These estimates took into account a dampening effect on domestic wheat prices as wheat for export was routed into the domestic consumption market, animal feed outlets, and inventory. In the impact analysis accompanying the final Karnal bunt regulations on compensation, USDA (1997) concluded that:

"...our quarantine measures were appropriate and justifiable when compared with the magnitude of the benefits achieved. Even a 10-percent reduction in wheat exports would have a significant effect on wheat sector income. It is estimated that a 10-percent decline in wheat exports would cause a decline in wheat sector of over $500 million."

USDA emphasized the 10-percent loss of export markets in its assessment because substitution and arbitrage opportunities made it unlikely that a Karnal bunt outbreak would lead to more than this amount of trade being diverted from countries imposing restrictions because of the disease.

In reviewing the USDA argument, Glauber and Narrod note that the original impact analysis failed to consider changes in consumer welfare resulting from lower domestic prices if wheat were diverted from export markets. Using observed price and domestic demand levels, and applying the domestic demand elasticity assumed by USDA, they estimate consumer surplus effects for each level of wheat export diversion. The resulting estimated annual effects on net welfare measured by consumer plus producer surplus ranged from $261 million for a 10-percent loss in exports to $976 million assuming a 50-percent reduction in exports.

Glauber and Narrod also conclude that the effects due to an outbreak of Karnal bunt outside the quarantined area should be evaluated on the basis of net present value of annual losses over the full period in which an outbreak is expected to have an adverse effect. Using an eight-percent discount rate and assuming losses sustained for ten years, they estimate that the discounted welfare effects ranged from just over $2 billion ($2.016 billion) for a 10-percent loss of exports to nearly $7.5 billion for a 50-percent loss.

In the original regulatory analysis, USDA estimated that the costs of the Karnal bunt regulations in 1996 incurred by producers, handlers, and other affected parties would be $44 million. These costs arose from six requirements: 1) plowdown of fields planted with infected seed ($1.2 million); 2) diversion of infected grain to animal feed ($4.2 million); 3) cleaning and disinfecting of railcars ($0.6 million); 4) loss of seed value ($6.0 million); 5) loss of value of negative-tested grain from millfeed treatment costs ($28.0 million); and 6) other assorted costs ($4.1 million).

Benefit-cost analysis for the eight quarantine options can be completed under the assumptions given above. For the baseline (option 1), the costs of destroying crops planted with contaminated seed and diverting positive-tested wheat to feed markets and is $5.4 million. The probability of an outbreak outside the quarantine area was reduced from certainty with no protocol to 0.0549. For a 10-percent diversion of exports with present value of costs of $2.016 billion, the expected loss due to an outbreak of Karnal bunt outside of the quarantined area is $110.7 million, and the welfare gain from utilizing the baseline option is $1.905 billion dollars. Each of the other quarantine options also shows a large expected benefit-cost ratio.

A large expected benefit-cost ratio does not imply, however, that each option is an economically efficient quarantine policy. Three options (3: restrictions on seed movements; 4: millfeed treatment; and 7: restrictions on seed movements and millfeed treatment) achieve little risk reduction compared to the baseline (option 1). Four other options (2: railcar cleaning; 5: railcar cleaning and restriction on seed movement; 6: railcar cleaning and millfeed treatment; and 8: railcar cleaning, restrictions on seed movement, and millfeed treatment) are more efficient policies in providing expected benefits for a given level of outlays. These options, along with the baseline (option 1), lie on or near an expected benefit-cost frontier, as shown in figure 3. Options 3, 4 and 7 are economically inefficient. Greater levels of risk reduction and expected benefits can be achieved with lower cost by other choices.

The results in figure 3 also show that most of the expected benefits of any quarantine procedure are achieved by the baseline (option 1). Using additional protocols adds to quarantine costs, but adds little to expected benefits. Consider the efficient baseline (option 1) and options 2, 5 and 8—in which the three protocols of railcar cleaning, restrictions on seed movement, and millfeed treatment are added sequentially. The expected benefits of each option, the marginal cost of each added protocol, and the expected marginal benefits of that protocol, at the mean value of risk and the more conservative 95^th percentile of risk, are shown in table 4. The results demonstrate that the use of railcar cleaning (option 2) provides $104 million in additional benefits for additional costs of only $0.6 million. The addition of a protocol restricting the movement of negative-tested seed (option 5) imposes an additional cost of $6 million, while the welfare gain is only $2.9 million when evaluated at the mean probability estimate. The added seed protocol is shown to be marginally cost effective when evaluated using the more conservative 95^th percentile value for the risk of outbreak. Finally, the protocol of millfeed treatment (option 8) adds $28 million to quarantine costs at the margin, but has a marginal benefit that is less than $0.1 million.

As shown in table 3 (and figure 3), Options 2 and 6 achieve nearly identical reductions of risk, as do options 5 and 8. Within each pair there is an efficient choice: 2 dominates 6, and 8 is dominated by 5. If the very similar levels of risk associated with these two choice-pairs are both considered acceptable by decision makers, then option 2 emerges as the most efficient among all four options, and thus is the least trade distorting (including, in this case, trade in the internal market).

To summarize, Glauber and Narrod find that the original Karnal bunt regulatory impact analysis ignored the effects of the quarantine policies on consumers, and therefore tended to overestimate the benefits of the quarantine. The original analysis also failed to look at the expected marginal benefits and costs of various quarantine protocols. Had the expected marginal effects been considered in the quarantine decisions, it is likely that at least two of the more controversial protocols, seed restrictions and the millfeed requirement, would have received closer scrutiny and possibly been rejected. The use of restrictions on movement of positive-tested grain and seed, together with railcar cleaning, results in significant risk reduction in an efficient manner.

This paper highlights the potential for complementarity between science-based risk assessment and economic-based benefit-cost analysis in regulatory decision processes. In particular, we use two case studies to highlight a key point: that economic benefit-cost analysis ought to play an explicit role in decision-making about SPS regulations. The SPS agreement of the WTO does not require countries to take into account benefit-cost analysis in making regulatory decisions—doing so might be considered a "WTO-plus good regulatory practice," as opposed to one, as Donna Roberts has previously described it, that is "merely legally defensible" (Roberts, Orden, and Josling 1999). A WTO-plus approach to regulation is unlikely to bring objections from trade partners since the results are likely to open trade opportunities otherwise precluded. The WTO agreement does require that countries employ measures that are least trade restrictive. This requirement alone can push countries some distance toward economically sensible SPS regulatory decisions, since for any given objective in terms of achieving a specified level of risk, the least trade distorting policies are those that achieve the risk objective most economically efficiently.

The avocado case study demonstrates the possibilities for either convergence or divergence between the policy implications from risk assessment and benefit-cost analysis. The USDA decision that a systems approach to risk management kept pest-infestation risk low enabled regulators to partially ease a longstanding ban on avocado imports from Mexico. This was a move to a less trade distorting policy, and for this move there is convergence of decision making criteria, at least in the sense that the limited-trade decision is only estimated to have a net positive effect on expected national welfare when pest risk is relatively low. The benefit-cost analysis also suggests an increase in welfare from a free trade policy even if pest infestation occurs with certainty. Free trade has not been considered a viable decision in recent policy determination, but additional steps toward trade opening will be evaluated. There is simultaneity between the rule, pest risk, and expected welfare. The more trade is opened, the higher may be the infestation risk, but also because of more trade, the less sensitive are the expected net welfare gains to the level of risk. One can imagine additional moves toward less trade distorting policies that raise infestation risk (while still keeping that risk below a level deemed acceptable) and simultaneously yield substantial net economic benefits. For example, what is the risk associated with avocado imports into the entire eastern United States over the full year? Perhaps such a policy captures most of the expected economic gains from free trade with relatively little additional risk exposure.

The Karnal bunt analysis by Glauber and Narrod, allows evaluation of just such relationships between quarantine options. Among eight options, they show a limited approach that restricts the movement of positive-tested grain and seed together with railcar cleaning achieves much of the potential risk reduction and expected welfare gains for the least direct cost. By highlighting the marginal cost and benefit of each protocol, they identify the most efficient options, and hence the options that would be least trade distorting because they impose the least costs on producers within the quarantine area. Two of their option pairs achieve nearly identical reductions of risk. Within each pair, there is an efficient and an inefficient choice. Moreover, if the very similar levels of risk associated with these two choice-pairs are both considered acceptable by decision makers, then the single benchmark-railcar cleaning option emerges as being the most efficient among the four options. Benefit-cost analysis can not make risk management decisions for policy makers, as between the two similar risk levels in the Karnal bunt case. But benefit-cost analysis can help policy makers choose the efficient protocol if they are willing to consider such a trade-off, and understand the costs incurred, and for what ends, if they are not.

Carman, Hoy and Roberta Cook. 1996. "An Assessment of Potential Economic Impact of Mexican Avocado Imports on the California Industry." Presented at the I.S.H.S Economic Division Meetings, Rutgers University, August.

Evangelou, P, P. Kemere, and C. Miller. 1993. "Potential Economic Impacts of an Avocado Weevil Infestation in California." Unpublished paper, USDA/APHIS, August.

Firko, Michael J. 1995. "Importation of Avocado Fruit (Persea americana) from Mexico, Supplemental Pest Risk Assessment." BATS/PPQ/APHIS/USDA, May.

Garoyan, Leon. 1995. "Proposed Rule for the Importation of Fresh Hass Avocado Fruit Grown in Michoacan, Mexico." Report prepared or by the California Avocado Commission by Management Research Associates, Davis, California, August.

Glauber, Joe and Clare Narrod. 2000. "A Rational Risk Policy for Regulating Plant Diseases and Pests." Draft Paper Presented at the Western Economics Association Annual Meetings, Vancouver, BC, June 30.

Nyrop, Jan P. 1995. "A Critique of the Risk Management Analysis for Importation of Avocados from Mexico." Report prepared for the Florida Avocado and Lime Committee and presented at the public hearings on the avocado proposed rule, Washington D.C., August.

Orden, David and Eduardo Romano. 1996. "The Avocado Dispute and Other Technical Barriers to Agricultural Trade Under NAFTA." Invited paper presented at the conference on NAFTA and Agriculture: Is the Experiment Working, San Antonio, Texas, November.

Podleckis, Edward V. and Michael J. Firko. 1996a. "Karnal Bunt: Likelihood of Spread via Conveyances, Harvest Equipment and Wheat Shipments." U.S. Department of Agriculture, Animal and Plant Health Inspection Service. May 8.

Podleckis, Edward V. and Michael J. Firko. 1996b. "Karnal Bunt: Special Risk Assessment Addendum." U.S. Department of Agriculture, Animal and Plant Health Inspection Service, May 14.

Podleckis, Edward V. and Michael J. Firko. 1996c. "Karnal Bunt: Special Risk Assessment Addendum II." U.S. Department of Agriculture, Animal and Plant Health Inspection Service, May 22.

Podleckis, Edward V. and Michael J. Firko. 1996d. "Karnal Bunt: Special Risk Assessment Addendum III." U.S. Department of Agriculture, Animal and Plant Health Inspection Service, May 28.

Podleckis, Edward V. and Michael J. Firko. 1997. "Karnal Bunt: Likelihood of Spread with Proposed 1997 Program," in Bunts and Smuts of Wheat: an International Symposium (V.S. Malik and D.E. Mathre, editors). Ottawa: North American Plant Protection Organization, pp. 229-271.

Roberts, Donna and David Orden. 1996. "Determinants of Technical Barriers to Trade: The Case of U.S. Phytosanitary Restrictions on Mexican Avocados, 1972-1995," in Understanding Technical Barriers to Agricultural Trade (David Orden and Donna Roberts, editors), St. Paul, MN: International Agricultural Trade Research Consortium, Department of Applied Economics, University of Minnesota.

Roberts, Donna, Timothy Josling and David Orden. 1999. "A Framework for Analyzing Technical Trade Barriers in Agricultural Markets." Economic Research Service, U. S. Department of Agriculture, Technical Bulletin Number 1876, March.

United States Department of Agriculture. 1997. "Karnal Bunt Regulatory Flexibility Analysis and Regulatory Impact Analysis." Federal Register 7 CFR Part 301 Docket 96-016-20, pp. 24753-24765, May 6.

United States Department of Agriculture. 1995. "Notice of Proposed Rule on Importation of Hass Avocados." Federal Register 7 CFR Part 319, Docket 94-116-3, pp. 34832-34842, July 3.

World Trade Organization. 1994. The Results of the Uruguay Round of Multilateral Trade Negotiations: The Legal Texts. GATT Secretariat, Geneva, Switzerland.