POVERTY THE WORLD BANK REDUCTION AND ECONOMIC MANAGEMENT NETWORK (PREM) Economic Premise MARCH 2010 · Number 4 53311 How Much Do Agricultural Policies Restrict Trade? Comparing Trade Restrictiveness Indexes Kym Anderson and Johanna Croser 1 Recently the Bank has provided new indicators for monitoring the extent to which agricultural policies restrict inter- national trade in farm goods. They come from two studies with differing methodologies and data sources, and each provides less-than-perfect estimates. This note shows how and explains why the two indexes differ for some countries. Policy makers and analysts are often keen to know the extent methodology set out in Lloyd, Croser, and Anderson (2010). to which agricultural policies reduce international trade This is based on sectoral estimates of the nominal rate of as- flows, as an aid to prioritizing negotiating efforts and unilat- sistance to farmers and the consumer tax equivalent (NRA eral reform agendas. There are various indicators used for and CTE) of domestic and border policy measures that af- that purpose. The most common are nominal rates of assis- fect each country's agricultural trade. Those NRAs and tance to farmers and related consumer tax equivalents af- CTEs, provided by Anderson and Valenzuela (2008) are de- fecting the prices that domestic consumers pay for farm rived by comparing domestic prices with prices of like prod- products. These measure the extent to which domestic ucts at a country's border.2 prices exceed those at a country's border. An alternative in- In this paper, we compare the estimates of indices from dicator is to use scalar index numbers from the Anderson the Anderson and Croser (2009) country-level TRI estimates and Neary (1994, 2005) family of trade restrictiveness in- (AC), and the Kee, Nicita, and Olerreaga (2008) OTRI esti- dexes. These measures provide a single theoretically sound mates, both available on the World Bank website (KNO).3 indicator of the trade effects of different policy measures We explore how the two series complement each other, why that is directly comparable across time and countries. they differ, and how estimation of the trade restrictiveness of Drawing on the seminal theoretical work of Anderson and agricultural policy can be improved in the future. Neary, two recent World Bank studies have attempted to an- swer the question of how much agricultural policies restrict Complementary Estimates of Agricultural trade nationally, regionally and globally. Kee, Nicita, and Ol- Trade Restrictiveness Indexes erreaga (2009) estimate, among other indices, a single trade reduction index (called an Overall Trade Restrictiveness Figure 1 presents the TRI aggregate estimates by AC for the Index or OTRI in their paper) for 78 developed and devel- import-competing and exportables subsectors and the over- oping countries for a snapshot in time (a single year in the all agricultural sector from 1960 to 2004. For developing early or mid-2000s). Updates of these have been reported countries as a group, the trade restrictiveness of agricultural regularly in the World Bank's Global Monitoring Report. policy was slightly increasing until the 1990s. Thereafter, it Anderson and Croser (2009) provide alternative annual declined, mostly due to reductions in Africa and Asia. For estimates of a similar index (called a trade reduction index, high-income countries, the TRI time path was similar but or TRI) for the agricultural sector of 75 developed and de- the causes differ. The aggregate results for developing coun- veloping countries for the period 1955 to 2007, using a tries are driven by the exportables subsector, which has 1 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK www.worldbank.org/economicpremise been taxed, and the import-competing subsector, which is cultural subsector, with countries ranked according to the being protected but by less than in high-income countries. AC estimates. In both studies there is considerable diversity Policies in high-income countries, by contrast, support both in the country-level index estimates. In line with the results exporting and import-competing agricultural products and, in figure 1, all high-income and transition economies have even though they favor the latter much more heavily, the positive index estimates, indicating unsurprisingly that farm assistance to exporters somewhat offsets the antitrade bias policies in the import-competing sectors of these economies from the protection of import-competing products in terms were trade-reducing in that period. There is a high degree of impacts on those countries' aggregate volume of trade in of correlation between the estimated series in the two stud- farm products. This is reflected in figure 1a in a much ies for many countries, especially the European Union (EU) smaller TRI for high-income countries overall for agricul- countries and most of Central and Eastern Europe's transi- ture as compared with that for just the import-competing tion economies. (Note that the common KNO estimate of subsector. the OTRI for member countries of the EU as a whole--49 Figure 2 presents the country-level detail from the two percent--is allocated to each member country in figure 2.) studies for 2000­04 (for which there are 49 countries in The differences between the two sets of estimates are common), showing the KNO estimates for the agricultural most noticeable at the top and bottom of figure 2(a). For sector OTRI based on import tariffs and NTMs alongside the EFTA countries (Switzerland, Iceland, and Norway) and the AC estimates of the TRI for the import-competing agri- Japan--countries with a strong comparative disadvantage in agricultural products--the AC estimates are much higher than the KNO estimates; while for Australia, the United Figure 1. Trade Reduction Indexes for the Agricultural Sector's States, and New Zealand--countries with a strong compar- Import-Competing and Exportables Subsector and Overall, All ative advantage in farm products--the AC estimates are Covered Tradable Farm Products, 1960­2007 much smaller than the KNO estimates. Figure 2(b) presents the estimates for those developing (a) High-income countries countries present in both data sets. Most countries had poli- 80 cies that were overall trade-reducing in the time period shown. For a few developing countries, the TRI estimate by 60 AC is negative, indicating that their agricultural policies in 40 aggregate were implicitly subsidizing imports slightly. The AC estimates for developing countries are generally smaller 20 than the KNO estimates. This tendency holds across the 0 three main developing country regions (Africa, Asia, and Latin America). There are only a few developing countries ­20 for which the KNO estimate is lower than the AC estimate, 1960­64 1970­74 1980­84 1990­94 2000­04 most noticeably Ghana and Sri Lanka. year These results are complementary. The AC estimates, based on historical data, enable greater insights into the re- (b) Developing countries strictiveness of policy over time. Also, the AC estimates for 50 import-competing and exportable subsectors give a stronger 40 indication of the antitrade policy stance in many countries, especially in previous decades, than is obtainable by exam- 30 ining indexes for just the import-competing industries. 20 However, the KNO series has the benefit that it can be read- ily updated from published secondary data. 10 0 Why the Two Studies' Estimates Differ 1960­64 1970­74 1980­84 1990­94 2000­04 year There are at least five reasons why the KNO and AC esti- import-competing exportables all covered tradables mates could differ. The most obvious empirical reason for the series to differ is that distortions data are drawn from differ- ent sources. In the KNO study, the main source is the WTO Source: Anderson and Croser 2009. Note: Regional aggregates are weighted using the average of the value of Integrated Data Base and UNCTAD's TRAINS database, production and consumption at undistorted prices. supplemented by WTO national Trade Policy Review reports. 2 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK www.worldbank.org/economicpremise Figure 2. Trade Reduction Indexes for the Agricultural Sector's Import-Competing Subsector, Selected Focus Countries, 2000­04 (percent) a. High-income and transition economies b. Developing economies Colombia Switzerland Ghana Iceland Sudan Norway India Japan Malaysia Ireland Philippines Romania Côte d'Ivoire Sweden Nicaragua United Kingdom Indonesia France Mexico Finland Kenya Italy Uganda Turkey Sri Lanka Denmark Tanzania Netherlands Chile Austria China Germany Madagascar Portugal Thailand Canada Brazil Spain Senegal Ukraine Bangladesh Russian Federation Egypt New Zealand South Africa United States Zambia Australia Nigeria ­20 30 80 130 180 ­20 0 20 40 60 80 Kee et al. OTRI, tariffs, and NTMs Kee et al. OTRI, tariffs, and NTMs Anderson and Croser TRI, import-competing subsector Anderson and Croser TRI, import-competing subsector Sources: Anderson and Croser 2009; Kee, Nicita and Olarreaga 2008. Note: The Kee, Nicita, and Olarreaga estimate for each country is for a single year in the mid-2000s for which the most recent data are available. Agricultural domestic support data (which are included in ures (positive or negative), plus an adjustment for the output- the KNO NTM estimate) are based on WTO members' no- price equivalent of direct interventions in farm input markets. tifications during the period 1995­98. By contrast, the data Where multiple exchange rates operate, an estimate of the im- used in the AC estimates are obtained from the World Bank's port or export tax equivalents of that distortion are included new Distortions to Agricultural Incentives database, which as well. The domestic-to-border price ratio is an appropriate provides price-equivalent distortion estimates for the pro- measure for the TRI analysis since it captures agricultural price duction and consumption sides of each commodity market and trade policies by comparing like products at the same based on direct price comparisons. By calculating domestic- point in the value chain, namely, the farm-gate level. to-border price ratios, the estimates include the price effects The different sources of data (and their different years), and of all tariff and NTMs plus any domestic price support meas- the way they are used, can potentially explain some of the dif- 3 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK www.worldbank.org/economicpremise ference in the estimates. The KNO estimates of their OTRI torted by import restrictions alone (the KNO approach) are higher than the TRI estimates by AC for agricultural- would give a higher estimate than a comparable TRI esti- exporting countries potentially because of the methodology mate by AC because the former would be based on data that used by KNO to capture the effects of NTMs. The KNO contain a fuller diversity of distortions across the import- method involves (1) estimating the restrictiveness of NTMs competing subsector.4 on import volumes by product and country, and (2) using The fourth reason why the two series could differ is the import demand elasticities to transform the estimated im- difference in the products included in the two studies. The port quantity to an ad valorem tariff equivalent measure. KNO estimates are based on a methodology where distor- The former step includes in the estimating equation a tions to import-competing products are weighted by ob- dummy variable for each NTM regardless of the extent of served import values (multiplied by import demand restrictiveness of that measure. For countries such as Aus- elasticities, as per the Anderson/Neary formulation of the tralia, the United States, and New Zealand, almost half of index). That is, the KNO estimates will only include prod- the OTRI estimates by KNO are due to NTMs. ucts for which there are nonzero imports at the HS six-digit The AC method of domestic-to-border price comparisons level, regardless of their importance to domestic production for like products at the farm-gate level of the value chain, or consumption. By contrast, the AC estimates are com- by contrast, provides an ad valorem equivalent directly. puted using a methodology where the weights are produc- While such measures based on price comparisons are likely tion and consumption based. Anderson and Valenzuela to be more accurate for covered products, there are many (2008) select agricultural products for inclusion in the data- food products imported for consumers that are not covered base because they are important contributors to the gross in the study because they were not important in domestic value of national production at undistorted prices, thereby production (see below). Also, generating such measures can minimizing the number of products needed to achieve the be computationally intensive, and updates are not yet as target coverage of 70 percent of that total value. The AC es- mainstreamed as the annual updates of UNCTAD's TRAINS timates are based on policy distortions to those 70 percent database. of products, including both import-competing and ex- The second reason to expect differences between the two portable subsectors. The TRIs are computed for the subsec- series is that the AC estimates are computed with the sim- tors separately as well as together; and they can be extended plifying assumption within each country that domestic price to include the nontradables subsector as well. elasticities of supply are equal across commodities, and the If the only difference between the two series was that same for domestic price elasticities of demand. That assump- KNO limit their sample to products facing actual import tion allows the AC estimates to be constructed by aggregat- competition, the AC estimates would give a more accurate ing distortions using as weights just the sectoral share of each indication of distortions to the domestic agricultural markets commodity's domestic value of consumption or production of a country because they include both import-competing at undistorted prices. The OTRI estimates, calculated with and export subsectors. However, the AC estimates could be a full set of country- and commodity-specified import de- improved by including more coverage of production and mand elasticities, has the benefit of capturing precisely the consumption beyond the current 70 percent level. At the differential responses of various commodity trades to a given same time the KNO methodology and OTRI estimates policy distortion. could be improved by including exportable products. The third reason to expect differences between the two The fifth and related reason for the difference between the series is that the KNO estimates are generated from a very two series is that the KNO estimates include only import- disaggregated data set (at the HS six-digit tariff line level, restricting policy distortions, whereas the AC estimates are which has more than 4,000 tariff lines) whereas the AC es- based on all distortions (positive and negative) to import- timates are based on a sample that averages just 15 farm competing and exportable industries. That set includes im- products per high-income country and 9 per developing port and export taxes and subsidies and ad valorem country (so as to cover around 70 percent of the gross value equivalents of nonprice border measures such as quantita- of each country's farm production). If the level of disaggre- tive trade restrictions or technical standards, the implicit gation had been the only difference between the two series, trade taxes associated with multiple exchange rates, as well the greater level of disaggregation in the KNO study would as domestic production or consumption taxes and subsidies result in more accurate TRI estimates. This is because the and the output subsidy equivalent of farm input subsidies KNO estimates correctly aggregate distortions from the net of input taxes.5 more disaggregated base, and the estimates reflect the full Differences in the estimated TRI series due to differing diversity of distortions across industries within the agricul- extents of product disaggregation, product coverage, and in- tural subsector under study. The OTRI for industries dis- strument coverage are evident from a comparison of the 4 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK www.worldbank.org/economicpremise KNO estimates with two alternative sets of AC estimates. manding protection from import competition--it will only The first comparison is between the KNO estimates and the include import restrictions and hence only a subset of dis- TRI estimates by AC for import-competing products in each tortions to agricultural trade (albeit probably the most dis- country (figure 3). Given the five differences between the tortive subset). For high-income and transitional economies, two series analyzed above, it is not possible to say a priori where almost all import distortions are protective, the AC whether the TRI estimates by AC should be larger or smaller estimates (with fewer sectors) are higher than KNO esti- than the KNO counterparts. For example, while the latter mates, most likely because the effect of including more will include many more products--including ones involving lightly protected products dominates. This could be partly little or no restriction because there is no local industry de- why temperate-climate countries such as Japan, Switzer- Figure 3. Trade Reduction Indexes for the Agricultural Import-Competing Subsector and for All Covered Tradable Farm Products, Selected Focus Countries, 2000­04 (percent) a. High-income and transition economies b. Developing economies Côte d'Ivoire Norway Sudan Japan Tanzania Ireland Zambia Sweden Philippines Finland Ghana Romania India Netherlands Indonesia Iceland Nicaragua United Kingdom Mexico Austria Senegal Germany Kenya Denmark Madagascar France Chile Italy Uganda Russian Federation Bangladesh Portugal Malaysia Spain Egypt Turkey Sri Lanka Canada Thailand Switzerland China Ukraine Brazil United States South Africa New Zealand Nigeria Australia Colombia ­20 30 80 130 180 ­20 0 20 40 60 80 Kee et al. OTRI, tariffs, and NTMs Kee et al. OTRI, tariffs, and NTMs Anderson and Croser TRI, all covered tradables Anderson and Croser TRI, all covered tradables Source: Anderson and Croser 2009. Note: The Kee, Nicita, and Olarreaga estimate for each country is for a single year in the mid-2000s for which the most recent data are available. 5 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK www.worldbank.org/economicpremise land, Norway, and Iceland, despite having highly protected compared to estimates for the other subsectors of agricul- import-competing agricultural sectors, have low OTRIs: ture (exportables and nontradables), thereby offering further many of their imports from tropical countries would face insight into the antitrade bias in different countries' policies. few if any restrictions (figure 3a). In African countries such The Kee, Nicita, and Olarreaga (2008) indexes, constructed as Zambia, where there have been import subsidies for sta- from a somewhat different methodology and data set, have ple foods, the TRI estimate is lower than the OTRI, suggest- the benefit of allowing for a comparison between the trade ing that the effect of including more policy instruments restrictiveness of agricultural and manufacturing import- dominates the explanation for the difference between esti- competing policies, offering insight into the extent of the mates (figure 3b). sectoral bias in protectionist national trade policies (usually The second comparison is between the KNO estimates favoring agriculture). The KNO estimates are more theoret- and AC's TRI estimates for all covered tradables (both ex- ically precise than the AC estimates because they are based portable and import-competing sectors). This brings the two on a more disaggregated data set and they capture the dif- series closer together in terms of product coverage, but the ferential responses of various commodity trades to a given AC estimates also include distortions to exportable indus- policy distortion through the inclusion of elasticity data. Be- tries. Once again it is not possible to say a priori whether the cause the KNO estimates are based on a routinely published AC estimates should be larger or smaller than the KNO es- data source, they can be regularly updated at low cost. timates. The extent to which the increased product coverage The level of disaggregation, the proportion of the sector in the AC estimates brings them closer to the KNO esti- included in the aggregation, and the types of policy instru- mates will depend on the type and extent of distortions to ments included in the analysis are all important determi- exportable versus import-competing subsectors. A compar- nants of indices of agricultural trade restrictiveness. The ison and figures 2 and 3 reveal that when exportable sub- more prevalent are NTMs, the more difficult it will be to sectors are included to generate a TRI for all agricultural avoid domestic-to-border price comparisons to get an accu- tradables, the TRI estimates generally are lower in 2000­04 rate measure. But such price comparison studies need to in- than those involving just import-competing subsectors. This clude not only products important in domestic production is because the exportable subsector tends to be less trade re- but also those important in domestic consumption but not stricted than the import-competing subsector. For example, be produced domestically (such as tropical products in tem- Switzerland and Iceland have large export subsidies in perate countries, and conversely). Such price comparison 2000­04 for several agricultural products (Josling 2009). studies are laborious and therefore expensive. Nonetheless, These trade-expanding subsidies reduce the TRI estimate they are being undertaken regularly by the OECD for grad- quite significantly when the exportable subsector is in- ually more and more countries, including for a large sample cluded. In contrast, Norway provides much lower assistance of African countries under a new joint project with the FAO to its exportable subsector than to its import-competing and national governments funded by the Bill and Melinda farmers, so the inclusion of exporting industries has a less Gates Foundation. Adding TRIs to the list of calculated in- significant effect on that country's TRI estimate (compare dicators by the OECD would enrich the policy analysis that the grey shaded bar for Norway in figures 2 and 3). As for will be possible with those estimates, without requiring any developing countries, Côte d'Ivoire, Tanzania, and Zambia more information that is currently needed to estimate each have trade-reducing policies in their exportable sub- NRAs/CTEs or PSE/CSEs if one is willing to adopt some sector, which leads to a higher TRI estimate for them in fig- restrictive assumptions about price elasticities. ure 3 than in figure 2. One final point: the TRI, with its inclusion of export sub- sectors, will be especially useful when assessing the restric- Conclusion tiveness of policy responses to spikes in international food prices, as in 2008 when many developing countries placed In recent years very considerable progress has been made in restrictions on exports of food. Efforts are currently under answering the question: how much do agricultural policies way to update the Anderson and Valenzuela (2008) and An- restrict trade? The two World Bank studies surveyed here derson and Croser (2009) databases to include that year. have approached the question from different angles, each producing complementary results as to the restrictiveness of Notes import-competing agriculture in developed and developing countries in the 2000s. The Anderson and Croser (2009) es- 1. Correspondence: kym.anderson@adelaide.edu.au. This timates have the benefit of being part of a longer time-series note is a product of a World Bank research project on Dis- of estimates, giving historical context to the current policy tortions to Agricultural Incentives (see www.worldbank.org/ position. The import-competing subsector estimates can be agdistortions). The authors are grateful for the distortion es- 6 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK www.worldbank.org/economicpremise timates provided by the authors of the various country case Research Project (www.worldbank.org/agdistortions) at the World studies, for funding from Trust Funds provided by the gov- Bank. Johanna Croser is a PhD candidate in the School of Eco- ernments of the Netherlands (BNPP), the United Kingdom nomics at the University of Adelaide, Australia. (DfID), and Ireland, as well as from the Australian Research Council. References 2. See also Anderson (2009) for a summary. The Ander- son and Croser (2009) database also contains estimates of a Anderson, J.E., and J.P. Neary. 1994. "Measuring the Restrictiveness of Trade Policy." World Bank Economic Review 8 (2): 151­69. TRI for the market of individual commodities, based on the ------. 2005. Measuring the Restrictiveness of International Trade Policy, methodology of Croser, Lloyd, and Anderson (2010). These Cambridge, MA: MIT Press. measures are novel because all previous work has focused Anderson, K. 2009. "Five Decades of Distortions to Agricultural Incentives." on constructing index numbers from the perspective of a Chapter 1 in K. Anderson, ed., Distortions to Agricultural Incentives: A single country. Perspective, 1955­2007. London: Palgrave Macmillan and Washington 3. The Kee et al. (2008) estimates are slightly different to DC: World Bank. Anderson, K., and J.L. Croser. 2009. National and Agricultural Trade and Wel- those in Kee et al. (2009), but we use the former because fare Reduction Indexes, 1955 to 2007. Database available at www.world they include a disaggregation of the OTRI into manufactur- bank.org/agdistortions. ing and agricultural subsectors of each national economy. Anderson, K., and E. Valenzuela. 2008. Estimates of Distortions to Agricul- 4. Another index reported in both the KNO paper and tural Incentives, 1955 to 2007. Database available at www.worldbank.org/ the AC database is a welfare reduction index, for which the agdistortions Croser, J.L., and K. Anderson. 2010. "Agricultural Distortions in Sub- variance of sectoral distortions is a component of the index. Saharan Africa: Trade and Welfare Indicators, 1961 to 2004." Mini- Space constraints preclude a discussion of welfare impact Symposium Paper for the Annual Conference of the Australian Agri- estimates. cultural and Resource Economics Society, Adelaide, 10­12 February. 5. As noted, the AC methodology can be extended to in- Croser, J.L., P.J. Lloyd, and K. Anderson. 2010. "How do Agricultural Policy clude domestic distortions to the nontradables subsector of Restrictions to Trade and Welfare Differ Across Commodities?" Amer- agriculture (Croser and Anderson 2010). By definition this ican Journal of Agricultural Economics 92 (forthcoming). Josling, T. 2009. "Distortions to Agricultural Incentives in Western Europe." subsector involves no trade distortions, so its inclusion in the Agricultural Distortions Working Paper no. 61, World Bank, Washington set of products necessarily will lower the sectoral TRI esti- DC. Available at www.worldbank.org/agdistortions. mates. Kee, H.L., A. Nicita, and M. Olarreaga. 2008. "Monitoring Report 2008-- Overall Trade Restrictiveness Indices." World Bank, Washington DC. About the Authors Available at http://go.worldbank.org/C5VQJIV3H0. Kee, H.L., A. Nicita, and M. Olarreaga. 2009. "Estimating Trade Restric- tiveness Indexes" Economic Journal 119 (534): 172­99. Kym Anderson is the George Gollin Professor of Economics and Lloyd, P.J., J.L. Croser, and K. Anderson. 2010. " Distortions to Agricultural formerly foundation Executive Director of the Centre for Interna- Markets: New Indicators of Trade and Welfare Impacts, 1960 to 2007." tional Economic Studies at the University of Adelaide, Australia. Review of Development Economics 14 (2), May (forthcoming). He is also a consultant and the leader of Agricultural Distortions The Economic Premise note series is intended to summarize good practices and key policy findings on topics related to economic policy. It is produced by the Poverty Reduction and Economic Management (PREM) Network Vice-Presidency of the World Bank. The views expressed here are those of the authors and do not necessarily reflect those of the World Bank. The notes are available at www.worldbank.org/economicpremise.