PAAP’s Electronic Newsletter
31 October 2008Volume 11 Number 21
MONITORING AND ASSESSING CURRENT AND FUTURE POLICY RESPONSES TO THE FOOD CRISIS
Given the breadth of possible policy responses to a food crisis in the longer term, the challenge of monitoring and assessing the impacts of such responses is great and hardly distinguishable from the need to monitor and evaluate policies and programmes to promote economic development and poverty reduction in general. This second and final part of the report by IFPRI on food crisis focuses mainly on methods of monitoring and assessing the short-term responses to a global food crisis and their impacts, although it also provides some discussion of the need for and approaches to assessing medium- and longer-term responses and impacts.
Introduction
N
ATIONAL governments have responded to the current world food crisis in several ways. Responsesintended to reduce the prices faced by domestic consumers in the short term, such asreductions in tariffs or consumption taxes, food price controls, actions against speculators andprofiteers, consumer subsidies, export bans or taxes, government food imports, and release of foodreserve stocks, have been quite common. Less common are interventions to increase domestic foodproduction by triggering a supply response to the crisis, such as through providing input subsidies orsupport prices for farmers. A third category of responses has sought to increase food availability foror the income of vulnerable or politically powerful groups through social protection programmes such asfood rations, food or cash for work, and cash transfer programmes.
The set ofpossible policy responses expands as the time frame forimpacts is increased. In the short term, policymakers cando little to change domestic food production if farmershave already made their planting and input use decisionsfor the upcoming harvest. Policy responses are limitedto changes in tariffs, taxes, consumer subsidies, foodprice controls, export restrictions, government foodimports, or the release of public reserve stocks. In themedium term, the scope of action widens--governments can implement price stabilization policiesbased on the use of reserves, tariffs, or subsidies;promote food production using subsidies, producerprice supports, or provision of agricultural supportservices; and extend social protection programmes. In the longer term, larger sustained impacts can be achievedthrough broader investments for economicdevelopment and poverty reduction.
Types of impacts expected frompolicy responses
As already noted, most of theshort-term policy responses aim to reduce consumerfood prices, which is a favorable effect from thestandpoint of net food buyers. Policies and programmes topromote increased food production also have beneficialimpacts on net food buyers to the extent that theyresult in reduced domestic food prices. Theycan also canbenefit food producers by reducing their costs ofproduction (such as through input subsidies)increasing producer prices (such as through pricesupport and producer subsidies), although the netimpact on producers depends on the relative strengthof these effects compared with the downward pressureon producer prices caused by increased production.
Targeted food aid or likely income-oriented interventionshave favorable impacts on the direct beneficiaries,and these may have beneficial spillover impacts on otherhouseholds or individuals such as by increasing demandfor goods and services provided by households as aresult of the increased incomes of beneficiaryhouseholds.All of these policy responses have costs andpotentially unfavourable impacts as well. All price-orientedinterventions, to the extent they are successful inreducing food prices, will reduce the incomes of net foodsellers and the incentive for producers to respond byincreasing production. Reducing tariffs or consumptiontaxes, increasing consumer or producer subsidies, orincreasing social protection programmes will have directbudgetary costs, potentially increasing governmentdeficits, credit shortages (if budget deficits are financed byborrowing), or inflationary pressures (if budget deficitsare financed through monetary expansion).
The benefitsof interventions may not be well targeted to poorer andmore vulnerable households, especially interventionsfocused on affecting market prices, leading to potentiallyhigh costs relative to the improvement in food securityachieved. Direct subsidies to producers or consumers orsocial protection programmes have more potential fortargeting, although targeting may not always be politicallyacceptable and may involve high administrative costs.Efforts to control prices and speculative behaviour maylead to black markets, and thus be ineffective, and maycontribute to corrupt practices by regulators. To theextent that price controls are effective, they can causeshortages that must be addressed by other rationingmechanisms, leading to inefficient and possibly inequitableallocation of commodities. Leakages and spillover effectsof interventions may also undermine their effectiveness.For example, export bans may lead to increasedcontraband food exports, while changes in public foodreserve stocks may be offset by induced changes inprivate stockholdings. Trade interventions such asexport bans on staple foods also can precipitate protectionistreactions by other nations, undermining the foodsecurity of trade-dependent countries as the internationalmarket for food becomes increasingly volatile.Many political, administrative, and economic conditioningfactors can influence the feasibility and impacts ofthese policy interventions. For example, the ability to usetariff or tax reductions to offset food price increasesdepends on the initial level of these tariffs or taxes andon the political will and fiscal capacity of the governmentto offset or forgo the revenues that would have beencollected. Budgetary constraints may also limit the use ofsubsidies, social protection programmes, or public sectorwage increases. The use of export restrictions dependson the government’s capacity to enforce such restrictions.
International treaty obligations under the WorldTrade Organization (WTO) or other trade agreementsmay also limit national governments’ ability to use trademeasures to buffer food price changes. The capacity ofthe government to enforce price controls or regulationson speculators will determine the effectiveness of suchmeasures, while the political context may promote orinhibit their use. The ability to expand the use of socialprotection programmes will depend upon prior experiencewith such measures and the administrative capacity of thegovernment to implement targeted approaches.These conditioning factors imply that low-incomecountries dependent on food and oil imports may belimited in their ability to use most of these potentialresponses effectively. Budgetary constraints are likely tolimit the use of large untargeted subsidy programmes,reductions in tariffs and taxes, or public sector wageincreases, whereas administrative capacity constraintswill often limit the ability to target social protectionprogrammes. Some social protection programmes, such asfood-for-work or cash-for-work, tend to be self-targetedand thus more readily usable than more administrativelycomplex approaches, such as conditional cash transferprogrammes. Low-income food-importing countries thatare large exporters of oil or other commodities whoseprices have also increased will have more budgetary capacity to use subsidies, tariff and tax reductions, orsocial protection programmes to buffer the impacts offood price increases, although they are still likely to facemany administrative capacity constraints. Higher-incomecountries tend to have more budgetary and administrativecapacities to implement a range of these options.
Monitoring and assessing theimpacts of policy responses
Monitoring involvescollecting data on selected indicators and observinghow those indicators change over time. The purpose ofmonitoring may be diagnostic or prescriptive. Forexample, changes in food security vulnerabilityindicators may be used to diagnose a serious problemoccurring for some population in some location,whereas such indicators combined with indicators ofthe conditioning factors affecting responses andoutcomes (such as indicators of prior investment andcoverage of social protection programmes) may help toprescribe promising policy responses.Monitoring by itself does not tell policymakerswhat impacts a given policy or programme is expected tohave (ex-ante assessment), is having (assessment duringimplementation), or has had (ex-post assessment).Toassess impacts, one must define the counterfactual orbaseline situation against which impacts are to beassessed and use analytical methods to measure thedifference in outcomes between the situation with thepolicy or programme being evaluated and the counterfactualsituation. Conceptually, the counterfactualsituation should be the situation that is expected tooccur (in an ex-ante assessment) or that would haveoccurred (in an assessment during implementation orex-post) without the intervention.
One of the main difficultiesin impact assessment work is that thecounterfactual situation is not observed (nor is thefactual situation—the situation with the policy—observed in ex-ante assessments).To address thisproblem, some assumptions and models, whetherexplicit or implicit, are necessary. For ex-anteassessments, predictive models are needed to predictwhat will happen with the intervention versus withoutthe intervention. Such models could be as simple asassuming that the quantities of the commodities ofinterest produced and consumed would be the same inthe future as in the recent past and that the onlydifference between the counterfactual and factualscenarios is in the prices of these commodities (for example, commodity prices might be assumed to bereduced by 25 percent due to a tariff reduction). Orthey could be complex multi-market or generalequilibrium models that seek to assess how changes inthe market for one commodity affect othercommodities and factors of production.Assessing impacts during or after an interventionoffers the advantage that the factual situation isobservable. The counterfactual situation, however, is not.Some common ways of addressing this problem are toassume that the situation observed before thepolicy or programme is what would have occurredwithout intervention (before-after comparison);
- Assume that the outcomes for some comparatorgroup not directly affected by the policy orprogramme represent the outcomes that would haveoccurred for the affected population without theprogramme (with-without comparison);
- Assume that the changes in outcomes for acomparator group represent the changes inoutcomes that would have occurred for theaffected population without the programme (double difference comparison); or
- Use a model to predict what would have happenedwithout the intervention.
The first approach (before-after comparison) isproblematic if other factors besides the policy orprogramme being assessed that affect the outcomes ofinterest are also changing over time. In this case, thebefore situation may be a very poor proxy for whatwould have occurred without the policy or programme.For example, food prices may be changing over time asa result of many supply- and demand-related factors, soattributing a change in food prices as due solely to apolicy change can be problematic.The second approach (with-without comparison) isproblematic if the comparison group is different fromthe affected group in ways that affect the outcomes of interest. This problem can be addressed by (1) randomlyassigning groups or individuals to “treatment” versus“control” groups using an experimental design, whichassures that the with and without groups are statisticallysimilar in all observable and unobservablecharacteristics; (2) selecting the comparison groups bymatching members between the groups on relevantobservable characteristics; or (3) using econometricapproaches to correct for selection bias. Because itensures that treatment and control groups are similar inboth observable and unobservable characteristics,random assignment is seen as the “gold standard”approach in impact evaluations of targeted programmes. However, this approach is not alwaysfeasible, politically acceptable, or appropriate to thenature of the intervention. Demand-drivendevelopment programmes, for example, do not easily lendthemselves to the use of supply-driven randomassignment.Moreover, all of these with-without comparisonapproaches assume that the comparison group isunaffected by the policy or programme being assessed. Forinterventions that affect food prices throughout acountry, it is difficult to find counterfactual householdsin the same country who are unaffected. Even fortargeted subsidies or social protection programmes,spillover effects of such programmes to non-participantsmay affect outcomes for potential comparison groups aswell. Such effects are easier to avoid for small pilotprogrammes or small targeted changes in such programmesthan for changes occurring on a large nationwide scale.
The third approach (double-difference) combinesthe strengths of the before-after and with-without
comparisons, since it nets out the effects of commonfactors affecting both with and without groups (like theeffects of common changes in prices affecting bothgroups) and fixed factors that may cause differences inoutcomes between the groups (like differences in theirabilities).This method is also subject to the shortcomingthat the comparison group may be indirectly affected bythe policy or programme, however, and to anymeasurement error problems in comparing differences.
The fourth approach (using a model to predict thecounterfactual) can overcome the problem of not beingable to identify a suitable comparison group unaffectedby the intervention. This approach is thus particularlyuseful for assessing the impacts of policies that affectfood prices and allows a similar approach to be used for assessing impacts ex-ante. The validity of theassessment will depend on the validity of the model andthe assumptions on which it is based, however, many ofwhich may not be readily testable.Nomethod of impact assessment is free of assumptions orpotential problems. The best method to use will dependon the type of policy or programme being assessed, thetime frame and outcomes of interest, the data availablefor the assessment, the ability to build on priorassessments and models, and the ability of key stakeholdersto use and comprehend the method used. The methods range from the simple to the complex, depending onthe nature of responses and impacts considered. Simplemethods require more restrictive assumptionsconcerning the responses of and impacts on producers,consumers, and others in the economy but are easier toimplement.
Price-oriented policies and interventions
Assessing the impacts of policies that affect householdsand individuals primarily through their impacts on foodprices—changes in tariffs, taxes, or subsidies; price orexport controls; grain reserve policies; and so on—canbe done in two steps. First, one estimates the impacts ofthese policies on domestic food prices and othernational-level outcomes such as fiscal and externalbalances. Second, the impacts of these price changes onhouseholds and individuals can be assessed using themethods discussed earlier for monitoring andassessing the impacts of food price increases. Becausethe methods and data needs for the second part of thisassessment were discussed earlier, this section focusesonly on the first part of the assessment.
A distinction can be made between policies thataffect food prices fairly directly, such as changes intariffs, and policies that affect prices by affecting thequantity of food available in the market, such as grainreserve policies. In the former case, a first approximationof the impact on prices is given by the policyitself. For example, a reduction in the tariff rate of US $10per tonne for an imported commodity can be expected toreduce the domestic price of that commodity by US $10per tonne if domestic markets are well integrated with theinternational market (perfect price transmission). If thisassumption holds, there is no need for any monitoringor assessment to determine the resulting domesticprice change. Because of transaction costs, market orgovernment imperfections, and other factors, however,the actual changes in domestic prices of the commodityin particular locations may be different (probablysmaller) than the change in the tariff rate. Hence it isuseful to monitor what happens to prices in differentlocations and assess the extent to which the policychange or other factors contributed to such changes.In the case of interventions affecting the supply offood in the market, such as release of public grainreserve stocks, a different approach is needed. The impacts of these quantitative supply shifts on prices must be estimated. This estimation can be done using a simplesingle-commodity partial equilibrium model, assumingthat only one commodity is affected, or using a morecomplex multi-market model, assuming that prices ofother commodities may also be affected. Other indirecteffects of the policy, such as the effects of releasingpublic stocks on private stockholding behaviour, may alsoneed to be taken into account to draw reliableconclusions about the impacts of the policy.
Reduction in import tariff:As noted, with perfectprice transmission to domestic markets, there is noneed for monitoring or analysis to know the impact of achange in tariff on the price of the affected commodity.The case of imperfect price transmission is thusconsidered here, along with methods for monitoringand assessing impacts on other commodities.To assess the potential short run impact of a tariffreduction ex-ante, considering imperfect price transmission,one multiplies the percentage import pricechange due to the tariff reduction by the elasticity ofprice transmission. For example, if a tariff reductionimplies a 10 percent reduction in the import price ofrice and the elasticity of price transmission for rice inthe country is 0.5, then the predicted impact on thedomestic rice price is a 5 percent reduction. This calculation assumes that the elasticity of price transmissioncan be estimated from available data or fromvalues in the literature and that the elasticity estimatedis valid for a change in the tariff. It is advisable,however, to use sensitivity analysis with a range ofvalues from the literature estimated for countries havingsimilar economic and policy environments. If suitablevalues are not available from the literature, the pricetransmission elasticity can be computed. Changing the tariff for one commodity may affectthe domestic demand, supply, and prices of othercommodities. For example, a reduction in the tariff onimported rice may reduce the demand for other staplecommodities, reducing their prices in the near term. The effects on supply may be mixed. Since the price of riceand other staples may fall, their production may also fallin the medium term, although substitution betweencrops in farmers’ supply decisions could lead toexpanded area of alternative crops.
A multi-marketsupply and demand model could be used to assess thesecross-commodity effects. Implementing such modelswould require data on the supply, use, and prices ofother commodities that are substitutes or complementsto the directly affected commodity, as well asinformation on the elasticities of supply and demand ofthese commodities (including cross-price elasticities).The supply and use data are likely to be readily available(though not always reliable) in the national statistics ofthe Ministry of Agriculture or from international organizationssuch as the Food and Agriculture Organizationof the United Nations (FAO). Estimated elasticities ofsupply and demand may be available for the countryfrom the literature, although typically it is difficult to findestimates of cross-price elasticities. If sufficient data onproduction, use, and prices are available for the selectedcommodities, these parameters could be estimatedeconometrically. Otherwise, it may be necessary to usevalues based on similar countries andcommodities in the literature, combined with expertjudgment and sensitivity analysis of the results.Such a multi-market modeling approach could beused either ex-ante or ex-post to assess the impacts ofa tariff change. If used ex-post, some of the predictedimpacts could be tested against actual data and used toimprove the model and estimated impacts. For example,the predicted impacts on prices of substitutecommodities could be compared with actual pricechanges, and the deviations could be used to adjust themodel parameters to obtain better predictions.Implementing this approach would be a fairly intensiveresearch endeavor and not something that is likely to bereadily implementable by government agencies in mostdeveloping countries.