Rent-sharing in the multi-fibre arrangement : the case of Mexico

WORf(ING PAPtARS | henatona Trad Intemational Economics Department The World Bank September 1993 WPS 1 191 Rent-Sharing in the Multi-Fibre Arrangement The Case of Mexico Geoffrey J. Bannister Market power affects the distribution of quota rents in the market for Mexico's exports of apparel and textiles to the United States under the Multi-Fibre Arrangement. Although rents from quotas on apparel are probably small in the case of Mexico, a significant share goes to U.S. importers for such product groups as under- wear and woven shirts. Paicy RubW*ainypFdinawtehdfinpoi(wctlopom =f wwimn.thzcUWcMd= imig3a&ufld .U__h um lmdin-velapmeiaw.Th.papm, dsbuWtdbydbReuwchAdvy ff,cuyLwyinmu,flauhar.rdl =dyo avWwaWshWbnd uhed Imawy. nd,.inutpm,d e & Th.yeadd so hafdgW to the Wei Bouk. s Dmod dDiue, iu_n amm, c my d itisun,r e_osm  lnwafAonl Trd WPS 1 191 This paper- a product of the Intemational Trade Division, Intemational Economics Department- is part of a larger effort in the department to analyze the effects of the Multi-Fibre ArrAngement on developing countries. The study was funded by the Bank's Research Support Budget under research project "Licence Prices and Rent Sharing in the Multi-Fibre Arrangement" (RPO 676-69). Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Aban Daruwala room S7-042, extension 33713 (September 1993, 35 pages). Bamnister investigates market power and the Bannister tests the existence of perfect distribution of rents in the market for Mexico's markets and rent-sharing for six groups of exports of apparel to the United States under the Mexican apparel exports to the United States Multi-Fibre Arrangement (MFA). between 1981 and 1990: sweaters, trousers, men's coats, women's coats, woven shirts, and Conventional wisdom holds that voluntary underwear. export restraints, such as those under the MFA, are superior to other kinds of trade barriers There are consistent differences between the because they allow developing countries to unit value of U.S. production and the Mexico receive the scarcity rents from quantity restric- export f.o.b. price of apparel in the U.S. market tions. Recently a number of studies have ques- adjusted for tariffs and transport costs. The tioned this orthodoxy. adjusted price of Mexican exports is consistently below the price for U.S. production, which Erzarn, Krishna, and Tan (1991), in particu- suggests that rent-sharing may be taking place. lar, have pointed out that if market power exists only on the side of the importers, they can Using modifications of the methods of acquire some of the fixed rents resulting from Erzan, Krishna, and Tan (1991), Bannister tests quotas, in a form of "rent-sharing." altemative explanations for the price difference - differences in the composition of Mexican In Mexico's case, rents resulting from MFA exports and U.S. production, and differences in restrictions are probably small, since few of the the quality of Mexican exports and U.S. prod- quotas imposed are binding. And other institu- ucts. tional arrangements - such as production- sharing under HTS 9802 and a liberal quota The existence of differences in composition regirne for goods made with U.S. inputs - between Mexican exports and U.S. production is further mitigate the MFA's restrictiveness. rejected for three of the six groups. Bannister also controls for the existence of significant Mexican exporters probably receive only a quality differences. fraction of available rents, says Bannister. The welfare implications of MFA restrictions, and of The results indicate that rent-sharing may market imrperfecdons that might lead to rent- exist for woven shirts and underwear (two of the sharing, are thus not as significant in Mexico as three groups in the samnple that are consistently they might be in countries for which conditions quota-bound). U.S. importers may receive up to are more restrictive. But even for the few rents 4'I oercent of available rents. generated in MeAico's case, some rent-sharing is taking place. The Policy Research Working Paper Series disseminates the findings of work under way in the Bank. An objective of the series is to get these findings out quickly, even if presentations are less than fuly polished. The findings, interpretations, and conclusions in these papers do i.it necessarily represent official Bank policy. Produced by the Policy Research Dissemination Center  Rent Sharing in the Multi-Fibre Arrangement: The Case of Mexico by Geoffrey J. Bannister The World Bank, Washington, D.C.  Rent Sharing in the Multi-Fibre Arrangement: The Case ot Mexico Geolfrey J. Bannisters The World Bank, Washington, D.C. .itroduction ..I 1. Institutional Arrangements ....................5 1.1 Export Licensing .................5,.......,....5 1.2 Trade Regimes ..................5 2. Testing for Rent-Sharing .............................. , , , , , , . .. 10 2.1 The Data ... . . . .10 2.2 The Price Differences ... . . .. . .. 11 2 .3 Testing Differences in Composition .....................17........ 17 2.4 Testing for Fixed Differences in Quality .............................. , .,.,. 23 Conclusions ........... 29 References .............30 Appendix: Data Sources and Calculations .............................,.,,.,.,,.31 *This is part of a larger project at the World Bank to examine the effects of quota restriction in the MFA on developing countries. I would like to thank Kala Krishna and Will Martin for helpful comments on ar. earlier draft.  Introduction Voluntary export restraints (VERs) are widespread in world trade in textiles and apparel. About 80 percent of world trade in textile and clothing, chiefly from developing to developed countries, is restricted by voluntary export restraints through various bilateral agreements which constitute the Multi-Fibre Arrangement (MF'A). The MFA allows industrialized countries to negotiate quantitative restrictions on specific toxtile and clothing items with individual developing country exporters. Exporting countries administer the quantitative restrictions, and, according to the conventional wisdom, reap the benefits of restricted supply in the importing countries thrcugh higher prices. A number of studies have recently questioned whether such benefits are really forthcoming. One approach points out the considerable welfare costs for exporters that may arise from the imposition of VERs.1 A second approach, followed in this paper, looks at the implications of imposing VERs in imperfectly competitive markets. The customary view of VERs assumes perfect competition in both the exporter's market for quota licenses, and the importer's market for each developing country's export product, with monopoly power existing only at the level of the country as a whole. The main result of these assumptions is that prices are set by the market, both for export licenses and for VER-restricted products, but are exogenous to the exporters and importers. With perfect competition in the quota license market, the level of the rent (the q'iota premium), is also given exogenously to economic agents. The assumption of perfect competition thus simplifies the analysis considerably by ignoring the institutional details of market structure and quota administration IDe Melo and Winters (1990), for example, point out that there are considerable welfare costs for exporters that arise from the imposition of VERs. The first cost is the efficiency loss due to government intervention in quota allocation rather than allowing for a straight auction of export rights. Even more significant, however, is the loss from the misallocation of resources that results in production and factor markets, even in the absence of government intervent;on in the quota allocation mechanism. The contractionary pressure on the industry subject to a VER can force factors out of industries in which they are most productive, and hence lead to significant efficiency losses for exporters. These losses must be balanced against the rent transfers from abroad. An empirical evaluation of VERs on Taiwan exports of footwear to the US. by Hamilton, De Melo and Winters (1992) indicates that they result in a net welfare loss for exporters in spite of the rent transfer. Martin and Suphachalasai (1990) also point to potential losses from depressed prices in residual markets in the context of the MFA.  Mexico MFA 2 that affect the market price under imperfect competition. When we allow for imperfect competition, however, the institutional details of price setting become essential to any investigation of the quota restrictions. Allowing for the possibility of market power, either in the market for export licenses on the supply side or on the part of US. retaile:s on the demand Fide, has implications for the existence and distribution of rents that accrue from the MFA between exporters and importers. The theoretical aspects of imperfect markets in quota all3cation and the effects of VERs in imperfectly competitive markets have been studied by Krishna (1990, 1992), Krugman and Helpman (1989), and empirically investigated in Aw (1992), Erzan, Krishna and Tan (1991, 1992) and Krishna and Tan (1992). Erzan, Krishna and Tan describe the implications of imperfect competition for the distribution of reints under the MFA. They make the crucial distinction between "rent-appropriation" and 'rent-sharing." When exporters have market power, they can set product prices to appropriate scarcity rents that accrue from quota restrictions. In this case, the size of the quota premium, (equal to the value of a license to export one unit of the quota restricted good) is det;rmined by the exporLer. Erzan, Krishna and Tan call this rent-appropriation. In contrast, when market power is on the buyer's side, the quota premium is a result of conditions in the quota restricted market, and is exogenous for both the exporter and the importer. The importer can then use market power only to extract some share of the fixed quote -ent. 'Vhen this occurs, the value of a license to export one unit of the quota constrained good will fall short of the quota premium, and hence only a fraction of the possible quota rent will accrue to the exporter. This is referred to as rent-sharing. Erzan, Krishna and Tan (1991) devise econometric tests for the presence of rent-sharing under imperfect competition, and apply them to Hong Kong exports of clothing to the US. restmcted under the MFA. They estim 3 the extent of rent-sharing, and find the potential rent was split unevenly between the U.S. and Hong Kong, with the U.S. share ranging from 48 per cent for skirts to 94 per cent for play suits. This paper investigates the case of Mexico's exports of apparel to the U.S. under the MFA. As the fifth most important exporter of apparel into the U.S. market, Mexico should be part of any study that measures the distribution of rents accruing from the MFA. Because of the institutional arrangements that govern trade between the U.S. and Mexico, however, Mexico presents a particular  Mexico MFA 3 chIlenge. First, there is no organized market for export licenses in Mexico, and hence no explicit export license price.2 Quota constrained exports are allocated to potpntial exporters on a historical basis. The implication of this arrangement is that there may be a high degree of concentration in the quota allocation for those products in which the quota is binding. Second, Mexican exports of apparel to the U.S. are influenced by market sharing arrangements that allow assembly of apparel in Mexico using U.S. inputs for: ,-export to the U.S. This trade may be within the same firm with operations in both the U.S. and Mexico, or may be the result of subcontracting arrangements between U.S. firms and Mexican assembly operations. Either of these ariangements has implications for export pricing strategies and market power relationships that affect rent-sharing. Third, even the quotas negotiated under the MFA have liberal provisions under an arrangement which essentially eliminates the restrictiveness of the MFA for Mexican exports produced with U.S. inputs. This arrangement, known as the Special Regime, went into effect in 1989, and clearly has some influence on whether quotas are binding and whether rents exist. Finally, it should also be noted that very few of the quotas imposed on Mexican exports are actually binding, so thLt even excluding the effects of the institutional arrangements described above, the potential rents accruing from MFA export restraints are probably very small. Thus, the welfare effects of quota restrictions on Mexican exports, and of market imperfections that might lead to rent sharing, are probably not large. Nevertheless, this case provides a suitable test for the existence of perfect markets and rent-sharing, with some bearing for a more sober assessment of the effects of MFA restrictions on developing countries. The first section of this paper contains a description of the institutional arrangements that affect Mexican exports of apparel to the United States under the MFA and their implications for the existence and distribution of rents. The second section investigates the existence of rent-sharing more directly by comparing the price of Mexican apparel exports to the United States under the MFA with U.S. prices for similar g-:ds. Under the assumption of perfectly competitive output markets, arbitrage in the. iporter's n', (i.e. in the U.S.) will lead to these two prices being cqual 2Thc transfer of export rights is illegal, and there is no evidence of informal mnarkets for these.  Mexico MFA t (assuming that impons are sufficiently close substitutes for home production), since the Mexican price of exports includes the implicit valu of a license to export the quota restricted product. Any difference between the two prices could be an indication of rent-sharing. The data indicate that Mexican export prices are consistently below the price of U.S. production. However, there are .0ternative explanations for these price differences. Chief among these is the difference in the composition of the groups of Mexican exports and U.S. production, or differences in the quality of the two products. Subsequent sections test for these differences and find that only in three of the six groups of apparel tested can differences in composition be definitively ruled out. Rent sharing estimates for two remaining groups that are quoa bound indicate that, when differences in quality are accounted for, U.S. retailers may have received up to 49 cents of every dollar in rents.  Mexico MFA S 1. Institutional Arrangements 1.1 Export l3censing Textile and apparel exports from Mexico to the U.S. under the MFA are restricted by visas which allow Mexico exports to pass through U.S. custonms. They are not restricted at the point of exit by Mexican customs officials. The export visas aLe distributed by Mexican officials according tc the amount of quota negotiated under the MFA. First prioritv to fill the quota is given to firms that exported in the previous year. Any left over quota is distributed to new entrants in the market. After the initial allocation, firms can increase their share of quota when there is unused quota quantity available, and when tne firm can show that it has alreadv exported 70 percent of -ts initial allocation. Visas are specific to a particular consignment, and are valid from the date of ibsuance until December 31 of the same year. What are the implications of this arrangement? The fact that the initial allocation of export permits (visas) does not take place through an auction mechanism meals tiat any rent created by the export restrictions benefits the exporter rather than the government. further, since visas are distributed on a historical basis it is probable that significant concentration exists in the distribution of these rents. Although there is no market for export permits, there is an unobservable iniplicit valuation the exporter confers on the export visa, which is equal to the amount he/she would be willing to pay for the right to export. To the extent that quotas are bindIng, this shadow price is positive, although it will vary from firm to firm. If quotas are not binding, the implicit price is zero. For a binding quota, this implicit price will be equal to the quota premium or the unit rent created by the quantity restrictions, and will be included in the F.O.B. price of Mexican exports to the U.S. that are restricted under the MFA. 1.2 Trade Regimes Two different institutional arrangements operate to diminish the trade barriers to Mexican apparel exports entering the U.S. The first of these is the provision under chapter 9802.00 of the harmonized tariff schedule (HTS 9802.00) that allows for special treatment of goods assembled in  Mexico MFA 6 Mexico from U.S. components. This provision allows American apparel firms to export cut cloth fcr assembly in Mexico's in-bond industry (the maquiladora industry) and re-import the final goods, paying tariffs only on the value added in Mexico. In most cases, the facilities in Mexico used for assembly are owned by U.S. firms, so that Mexican apparel exports entering the U.S. under HTS 9802.00 consticute movements of goods within the same firm, even though they Are registered as imperts into the U.S. under the MFA (USITC, 1991). HTS 9802.00 is essentially a tariff provision, but it has implications for rent sharing. If rents exist, then U.S. firms with assembly operations in Mexico owning a large snare of the quota will be the principal beneficiaries of the quota restrictions.3 The evidence in Table 1 shows that for Mexican apparel exrports to the U.S. this is probably the case. From 1988 to 1991 between 85 and 90 percent of all Mexican appare' exports under the MFA entered the U.S. under HTS 9802.00 provisions, with between 60 and 70 percent of their value added in the U.S. Table 1. Percentage of MFA Exports from Mexico to the U.S. Entering Under HTS 9802.00 1988 1989 1990 1991 U.S. U.S. U.S. "J.S. 9802.00 Value 9802.00 Value 9802.00 Value 9802.00 Value Added Adided Added Added All MFA 72.2 69.3 76.87 61.5 73.2 65.7 75.5 66.2 Apparl 84.9 69.4 90.68 61.5 88.32 66.3 89.52 66.3 Other 28.3 68.9 30.08 61.5 28.01 63.5 29.27 66.01 Source: U.S. Department of Commerce, Office of Textiles. The second institutional arrangement affecting Mexican exports of apparel into the U.S. is a special provision of the MFA known as the Special Regime. It effectively eliminates quotas under the MFA for apparel assembled in Mexico from fabric cut and formed in the U.S. The test for eligibility for the Special Regime quota treatment is more stringent than that for HTS 9802.00 tariff provisions 3This is not a case particular to Mexico. Large U.S. importers that also manufacture and assemble appaiel overseas, such as Liz Claiborne, often have claims to large fractions of available quota in developing countries.  Mexico MFA 7 since the former requires that fabric be formed and cut in the U.S. while the latter only requires the cutting to take place in the U.S. Although Special Regime quotas a, e often filled, they are administered in such a way that utilization rates have been allowed to exceed 100 percent. Thus it may appear that rents exist for those exports that entei the U.S. under the Special regime vhen, in fact, no rents are being generated. The special regime was instituted in 1988, although it only became effective in 1989. In 1990, 43.5 percen; of apparel exports fromn Me- ico into the U.S. under the MFA entered under the Special Regime.4 In addition to the special regime, there are different quota arrangements within the MFA that affect the restrictiveness of tht quotas. rhere are three types uf quotas: specific limits, desigi .ed levels, and consultation mechanisms. A specific limit quota is a quantity constraint which increases at a fixed rate per year (in most cases six percent, with the exception of cotton fiber which increases at two percent per year). If the quantity restriction is met, then a specific limit quota is binding. A designated level is an informal barrier whose restrictiveness depends oin the discretion of the administrators of the quota in the United States. Mexican officials can request an increase in the quota for a specific year which may or may not be granted depending on the 'J.S. administrator's judgment as to what effect this will have on the U.S. market.5 Finally, consultation mechanisms inpose no quantitative limit, but establish a mechanism by which the U.S. can consult with Mexico when exports are perceived to be affecting the U.S. market adversely. In practice, these consultat;on mechanisms have not been binding. The arrangements described above affect a very large proportion of Mexican exports of apparel to the United States. Table 2 shows how " ay affect the groups of apparel that have been most bound by quota arrangements between 1981 and 1990.6 The first column shows the average rate of quota utili7ation (quantity of exports/quota). In general, a consistent quota utilization rate of 90 4For a more detailed description of these arrangements and their implications for textile trade under the North American Free Trade Agreement, see Bannister and Low, (1992). 5According to officials from the Department of Commerce, there are some MFA categories in which petitions for increase of the design'ted levels have been denied, so that there is reason to believe that at least in some cases these quotas are binding also. 6The composition of these groups is discussed in the Appendix.  Mexico MFA 8 percent or above may be considered to indicate that the quota is binding. The groups in Table 2 are aggregations of M.FA categories, some of which were quota bound and some of which were not. In addition, not all groups were bound in all years. The most consistently bound groups were trousers, woven sh'rts and linderwear. Of the groups presented only sweaters did not have a significant component of exports entering the U.S. under HTS 9802,00; between 80 and 90 percent of the exports of all other groups entered under HTS 9802.00, with an average value added in Mexico between 43 percent for shirts and 22 percent for unierwear. This suggests that some of the rents accruing from the MFA in these producL. are captured by U.S. firms assembling apparel in wMexico. Under type of quota, all w re subject to either designated consultation levels or specific limits, except for sweaters, which were only subject to consultation mechanisms. Trousers, woven shirts, and underwear, being the most tightly constrained groups of exports, are probably where the rents, if any, are being generated.7 Yet, as shown in the last column of Table 2, a high proportion of these exports in these groups entered the U.S. under the Special Regime and thus are eftectively not bound, at least from 1989 on. Table 2 thus reinforces the contention that available rents from MFA quota restrictions are small, and under the most optimistic assumption that exporters receive all the rent, Mexican producers still capture only a portion. 7In the analysis that follows we consider these to be the quota bound groups.  Mexico MFA 9 Table 2. Summary data for Apparel Exports Average Average Average Average Group Quota Percent of Percent Type Percent Utilization MFA under Value of Special 9802 Added Quota Regime (1981-90) (1987-90) in U.S. (1988-90) 1. Sweaters 38.10 1.20 33.00 cm 00.00 2. Trousers 70.70 83.10 67.30 di/sI 83.70 3. Men's Coats 42.90 88.20 65.50 dl 4.60 4. Women's Coats 51.80 80.10 70.10 di/sl 29.20 5. Woven Shirts 63.40 89 70 57.70 sI 74.20 6. Underwear_ 59.30 82.30 _ 78.30 dl/sl 59.70 S,urce: World Bank data tapcs and U.S. Department of Commerce data. cm-consultation mechanism; dl-designated level; sl=specific limit.  Mexico iFA 10 2. Testing tor Rent-Sharing 2.1 The Data Thc data used for this study are similar to those used by Erzan, Krishna and Tan (EKT). Because of conflicting classification systems between U.S. data and MFA import data the categories used were aggregated in the same fashion as the EKT data, choosing groups of apparel that minimized the differences between U.S. production groupings and aggregate MFA category groupings. Six industry groups were examined: sweaters, trousers, men's coats, women's coats, woven shirts, and underwear.8 The data include observations for the following variables from 1981 to 1990, where j indicates the apparel group, and t indicates the year: pitUS= the unit value of U.S. production. pjtm = F.O.B. price of apparel imports from Mexico. tjt = Ad valorem tariff in the U.S. Tit = Unit transport cost from Mexico to the U.S. pitm = Adjusted Mexico price, where pjtm = pjtm (1 +tj ) + Tjt 9 QjUS = U.S. sales of U.S. production. Qitm e Mexican exports to the U.S. market. Hjt = Numbers equivalent of the Herfindahl index of concentration among Mexican exporters. Vit = Quota level for Mexican exports to the U.S. Uit = Quota utilization rate defined as Ui, = Qjtm /Vit The sources and composition of these data are found in the data appendix. tThe composition of these groupings in terms of MFA categories is explained in the appendix. 9A further adjustment was made to the unit value of Mexican exports to take into account the lower tariff rates levied on imports with a high proportion of U.S. content under the production sharing arrangements codified in HTS 9802.00.  Mexico MFA 11 2.2 The Price Differences As mentioned above, the implicit valuation of the quota rents is included in the Mexican F.O.B. price of exports to the U.S. Thus, we can test for rent-sharing by comparing the unit value of U.S. production with the Mexican F.O.B. price, appropriately adjusted for tariffs, transport costs, and HTS 9802 tariff concessions. Arbitrage in the U.S. market will cause these two prices to equalize if markets are competitive and all goods are homogeneous within each group. It is reasonable to assume that Mexican exporters are small, and therefore price-takers, in the U.S. market. However, if U.S. importers have monopsony power, this can lead to rent-sharing, if they can maintain a lower price for their imports than they pay for U.S. production. Chart 1 plots the adjusted Mexican F.O.B. price on the vertical axis against the U.S. price on the horizontal axis for all six groups of products examined in this study. The arrow represents the forty-five degree line. The chart clear!y shows that there is a significant difference between the two prices. The U.S. price is above the Mexican price in almost every instance, inaicating that either sustained quality or composition differences or rent-sharing may exist. To test the significance of this difference, the following regression was run using time series data from 1981 to 1990, pooled over the six apparel groups: (1) pjm = a + ppjtuS + yHjt + 8Ujt + fVjt + ejt At this point the variation over groups of apparel is not considered, and so a is maintained constant. As Erzan, Krishna, and Tan point out, in this equation, the right hand side variables can be considered exogenous to the Mexican exporter. If there is no rent sharing, and markets are perfectly competitive, and if all goods are assumed homogeneous within g,oups, then pj,m = pjtUs. In this case we expect the coefficient on the U.S. price not to be significantly different from one, and all other parameters not to be significantly different from zero. If, on the other hand, rent sharing exists, implying monopoly power on the part of U.S. importers, or the assumption of homogeneous goods is violated (by differences in quality or composition, for example), then we expect the coefficient on pul  Mexico MFA 12 to be different from unity, and the coefficients on the other variables to be significant. Hjt, the numbers equivalent of the Herfindahl index, is an indicator cf concentration in export license holdings among Mexican exporters.10 All other things being equal, an increase in concentration will afford Mexican exporters greater bargaining power vis-A-vis U.S. importers, allowing them to retain a higher share of the rent,11 In this case, we expect the coefficie.,: on the Herfindahl index to be positive. The quota level VjP, and the rate of quota utilization Ujt, reflect conditions in the supply of export licenses and the restrictiveness of the quota. All other things being equal, we expect an increase in the quota level to reduce the implicit export license price and hence for the coefficient on this tern to be negative. An increase in the utilization rate, on the other hand, will make export licenses more scarce, and hence increase their implicit price. Under the hypothesis of rent-sharing, we thus expect the coefficient on Uj, to be positive. 1I0Te numbers equivalent of the Herfindahl index is calculated as liZ si where si is firm i's share of license holdings. In our application the shares are weighted by the number of firms in each apparel group. See the Appendix for details. IlIn the extreme case where there is one monopsonist importer in the U.S. and all import licenses are held by one exporter in Mexico, Erzan, Krishna and Tan (1991) show that if the license price is determiined by the outcome of a Nash bargaining process, the license price is proportional to the level of the exporter's bargaining power.  Mexico MFA 13 Figure 1 Scatter Plot, All Apparel Groups Mexico (Prie per unit in U.S. dollars) 30.00 45 dog. 25.00 20.00 15.00 10.00 * ,E. 5.00- 0.00 I I I I 0.00 5.00 10.00 15.00 20.00 25.00 30.00 United States  Mexico MFA 14 Table 3a. Regression Results for Equation 1 - Linear Specirlcation. Independent All Groups Bound Groups Unbound Groups Variables (2,5,6) (1,3,4) Intercept 2.4515 A 0.6050 b 4.6597 a (0.5964) (0.3146) (0.9263) Pjtu' 0.4094 9 0.4676 A 0.4474 A .______________ (0.0257) (0.0310) (0.0478) ujt -0.8343 -0.2752 -2.483 c _______________________ (0.9247) (0.4502) (1.4850) Vjt -2.29 X 10-8 -6.79 X 10-9 6.21 X 10i7 (2 X 108) (I X lo-8) (7 X 10-7) Hijt -0.0012 0.0029 -0.0153 b (0.0022) (0.0011) (0.0075) Adj. R2 0.872 0.942 0.775 # observations 60 30 30 t Test, = _1 -22.89 reject ' -17.16 reject A -11.5 reject a F Test, ,=1 and 338.71 reject ' 292.09 reject & 268.81 reject ' Standard errors in parentheses. A Significant at 1 percent. b Significant at 5 percent. c Significant at 10 percent. Table 3a presents the results of running regression (1) on all groups, the quota bound groups and non-bound groups.12 The first significant result is that for all three regressions the hypothesis of perfect competition (3=1 and a=y=5=4=0) and the hypothesis that the coefficient on the U.S. price ,B is equal to one, are both rejected. While this does not confirm the existence of rent sharing, it is consistent with the rejection of the hypothesis of perfect competition in the market for Mexican export to the U.S. in these groups of apparel. The intercept and the coefficient on the U.S. price are significant in all three regressions. In this context, the intercept can be interpreted as refiecting the effects of any fixed difference between the two prices common to all groups of apparel (the "fixed component"), and the coefficient on the U.S. price reflects the change in the Mexican price for every 12Tbe definition of the groups is based on the data in Table 2.  Mexico MFA 15 unit change in the U.S. price, all other hings held equal (the 'marginal component"). 13 The fact that the intercept and the coefficient on the U.S. price are significant in all three regressions does not allow us to say anything about the probable cause of the price differences. It is interesting to not, however, that the fixed component on the bound group is significantly lower than the one on the unbound group, although the marginal components are similar in magnitude.14 One obvious possible interpretation for this is that at least part of the difference between the two prices is being captured by exporters in Mexico in the form of quota rents, reflected in a higher price of Mexican exports. This interpretation is supported by the coefficient on the numbers equivalent of the Herfindahl index, which is positive and significant for the bound group, but negative and significant for the unbound group. To test the robustness of the results in Table 3a, an alternative specification of the model in the logarithms of the variables was used. Under this specification, the coefficient on the log of the U.S. price can be interpreted as an elasticity of price transmission, while the other coefficients can be interpreted as elasticities reflecting the effects of the different characteristics of the quota-license market on the Mexican FOB price. The constant term can be interpreted as a proportional shift parameter. To determine whether the log specification is superior to the linear specification in (1), a Box-Cox test for model specification was used. 15 The test consists of comparing the sum of squared residuals of the two models after performing a simple transformation of the data. The sum of squared residuals for the linear specification was 4.163. For the log-linear specification it was 2.289. We thus can conclude that the log-linear specification fits the data better. The results of the regression in logs are presented in Table 3b. As before, the hypothesis of perfect competition (p = 1 and a;y=y-4=0) is rejected for all three regressions. However, it is 13Later in the paper we allow the intercept to vary across groups to explore the possibility that the fixed component can be explained by group-specific fixed differences in quality between U.S. production and imports from Mexico. 14Chow tests were run to see if the coefficients for the bound and the unbound groups were significantly different from each other. Tle tests rejected the null hypothesis that there was no difference between the groups at the one percent level. 15For a description of the Box-Cox test see Maddala (1992) p. 220 or Fombv et al. (1984) p. 423.  Mexico MFA 16 interesting to note that the hypothesis for P = 1 cannot be rejected for the unbound groups. In contrast with the linear specification, the intercept terms cease to be significantly different froin zero. However, the coefficient on the log of the U.S. price is significant at the one percent level for all groups, and the coefficient on the log of the utilization rate is significant at the fire and ten percent level. The log of the Herfindahl index is significant and negative at the five percent level for the unbound groups. Although it is difficult to interpret the negative coefficients on Uit and Hjt , these results are not inconsistent with those of the linear moecification, or with the hypothesis of rent- sharing for the bound groups of apparel. Table 3b. Regression Results for Equation 1 - Log-linear Specification. Independent All Groups Bound Groups Unbound Croups Variables (2,5,6) (1,3.4) Intercept 0.2639 0.3082 0.9294 (0.4178) (1.1384) (0.9968) log pjtus 0.8090 A 0.8027 a 0.8752 a (0.0491) (0.0450) (0.1209) log Ujt -0.0995 c -0.1003 c -0.1776 b (0.0660) (0.0735) (0.1012) log Vjt -0.0244 -0.0525 -0.0166 (0.0248) (0.0552) (0.0857) log Hjt -0.0211 0.0759 -0.2501 b (0.0583) (0.0674) (0.1293) Adj. R2 0.939 0.976 0.735 # observations 60 30 30 t Test, P= 1 -3.889 reject a 4.377 reject a -1.031 cannot reject ' F Test, P=1 and 97.093 reject a 97.096 reject ' 49.113 reject ' Standard errors in parentheses. a Significant at 1 percent. b Significant at 5 percent. C Significant at 10 percent.  Mexico MFA 17 2.3 Testing Differences in Composition Although the results in the previous section seem to indicate the possible existence of rent sharing, there are other porible explanations for the difference between the Mexican export price and the U.S. price. One possible explanation is a difference in the composition of the groups of apparel, with Mexican exports concentrating on the lower value MFA categories that make up the groups, and U.S. products concentrating on the high value end. As explained in the data appendix, each of the groups examined is an aggregation of categories in the MFA export data and different categories of U.S. production data. The problem faced when attempting to compare the unit value of U.S. production to the unit value of Mexican exports under the MFA is that at the most disaggregated levels the two category groupingc are not compatible. In aggregating Lhem to the more comparable group levels, some compositional bias is inevitable, and it may show up in the marginal rent-sharing parameter f3, the coefficient on pjtur, or in the intercept term. This section tests whether the price differences detected above can be explained entirely by this compositional bias. If composition bias cannot be ruled out, then the assumption of homogeneous product groups cannot be maintained and some accounting for product differences must be made. The procedure for testing compositional differences follows that of Erzan, Krishna, and Tan (1991), with some modifications to allow for a more efficient estimation of the parameters and to explicitly test the assumptions underlying the test procedure. The aggregate prices can be decomposed into their production weighted components as follows: (2) pu = piju. (Qiju/ = piju. wjus i,...,n, j1,...,6 where the subscript i refers to the category belonging to apparel group j. Qijus is the quantity of U.S. output in category i of group j, and QJUu is total output in group j. Hence wijUs is the quantity weight cF category i in group j of U.S. production. Similarly, for Mexico: (3) pjm (Qj/Qj) = i Wij, i= ,...,n, j=1,...,6. i I  Mexico MFA is The difference between the Mexican export F.O.B. price and the U.S. unit value of production at the group level can then be expressed: (4) pjm pjU = - ppjmw p im wpij us .wus Data only exist for pjm, pjus, pfm, axid wim. To get around this problem we assume that the following relationship holds for each sub-group (MFA category) i within each group j: (5) pjm = cj + p,jPiUs + P2jXjj + edj . where Xij is a general term for the independent variables included in equation (1): the level of the quota, the utilzation rate, and the Herfindahl index of concentration in export supply. Note that we assume a, Pl, and P2 are constant over all the members of each group j. Solving for pjus and substituting into (4) yields: (6) pjm - pjus = yj + E (wijm - sljwiiu)pijm + 82j wijuaxi, + 51j wijuseij . i I where yj = cj/plj , 8j = I/Plj , and 82j = P2j/Pjj . With one key assumption we can estimate equation (6) and test the coefficient on pijm to see if the composition effect is statistically significant. The assumption is that 82j = 0 for all j, that is, that the change in the Mexican export price is entirely determined by the change in the U.S. price, and variables such as quota levels, quota utilization, and concentration have no systematic effect. This seems to be a strong assumption, but it is borne out in the aggregate by the results of estimating equation (1) for all groups. Introducing this assumption, the regression equation then becomes: (6') pjm rpius = yj + upijm + uj where 00 = (wijm - 8iwiju), and uj = 8iIZwijuseiJ In order for E(uj) = 0 and the regression to be well specified, we have to make the additional assumption that the E(wije1j) = 0, that is , that the U.S. quantity weights and the error term are independent random variables.  Mexico MFA 19 To test for significant differences in the U.S. and Mexican category weights wUus and wijm, a compositional bias, it is necessary to impose the additional restriction that 8Si, the marginal component of the price relationship in each group, is equal to 1. 16 When this is the case, the sum of the coefficients on the p1m terms in equation (6') is equal to zero, since wijm = wiju = 1 . We can test the validity of this restriction using a joint F test on the coefficients of the pijm terms in (6'). If the test cannot reject the hypothesis that ejj = 0, then we take this to be sufficient evidence that i 81j = 1. At this point we can examine the significance of the individual 9ij coefficients for evidence of a composition effect, that is, a difference between wijm and wij". In addition, we can test the joint hypothesis that all the 90 coefficients are equal to zero for further evidence of a composition effect. If the hypothesis that