A comparative analysis of productivy in Brazilian and Mexican manufacturin industries Armênio de Souza Ranel and Fernando Garcia de Freas Abstract Tis article analyses productivy trends in Brazilian and Mexican manufacturin industries between 1995 and 2009, a period in wic international competion intensified sarply. A total of 14 manufacturin industries are considered, usin two metods based on: (i) te Leontief (1951) model to measure te consumption of intermediate oods used in production; and (ii) te analysis of total factor productivy (tfp). Te studies performed sow tat manufacturin trends ave divered in te two countries. In Mexico, an increased need for imported oods and services was offset by a reduction in domestic oods and service requirements, and an increase in te tfp of production. In te case of Brazil, te fact tat manufactured oods markets are more isolated from forein trade seems to ave contributed to a weak productivy performance. KEYWORDS Industry, industrial enterprises, manufactures, productivy, comparative analysis, input-output analysis, econometric models, Brazil JEL CLASSIFICATION C67, L60, O3, O40, O47 AUTHOrs Armênio de Souza Ranel, doctorate professor of te Scool of Communications and Arts (eca) of te Universy of São Paulo (usp), Brazil. armenio@usp.br Fernando Garcia de Freas, economic director of te Brazilian Camber of Services and economic adviser to te Brazilian Aluminium Association (abal). fernando.arcia.freas@mail.com
182 cepal review 115 april 2015 I Introduction Over te past two decades, te Mexican and Brazilian economies experienced profound transformations, larely driven by forein trade. Te reduction of import quotas, toeter w te elimination of non-tariff barriers and trade interation w neibourin countries, radically caned te structure of te two countries forein trade. In manufacturin industry, Brazil and Mexico suddenly faced external competion, particularly from East Asian countries. As noted by Mesqua (2007), te emerence of Cina on te world industrial stae posed major callenes to te Latin American economies, because te static and dynamic productivy differentials of Cinese manufacturers place enormous constraints on te productive potential of Brazilian and Mexican manufacturin industries. Followin a lenty rowt period, te sare of manufacturin industries in Brazilian and Mexican ross domestic product (dp) fell sarply. Accordin to te statistics and indicators database (cepalstat) of te Economic Commission for Latin America and te Caribbean (eclac), manufacturin-industry dp sares peaked in 1985 at 35.9% in Brazil, and in 1988 at 27% in Mexico. In 1996, manufacturin value added ad declined to just 19.6% of Mexican dp and 14.8% of Brazilian dp. Tis loss of dp sare as continued since, albe at a slower pace: in 2011, manufactured oods represented just 17.8% of Mexico s dp and 12.4% of Brazil s. 1 Tis result was due mainly to te slowdown in industrial rowt. Katz (2000) found tat manufacturin industry output rew by 3.8% per year in Mexico and by 2.8% per year in Brazil between 1970 and 1996, but rowt was slower in te period 1996-2009. Fiures from te World Input-Output Database (wiod, 2012) sow tat te annual rowt rate of manufacturin production fell to 1.2% in Brazil and 1.6% in Mexico, in tat period. Moreover, te slackenin of manufacturin productivy rowt was even more serious tan te decrease in s sare. Accordin to te study by Katz (2000), between 1970 and 1996, labour productivy Te autors are rateful to an anonymous referee and to professors Ana Lélia Manabosco and Roério Cesar de Souza for valuable comments and suestions made on previous versions of tis article. 1 For furter details, see Mesqua (2007). rose by 2.9% per year in Mexican manufacturin industry and by 1.9% per year in te same sector in Brazil. wiod (2012) data report an increase in value added per worker of just 0.1% per year in te Mexican manufacturin industry between 1996 and 2009, and a muc worse suation in Brazil, were value added per worker actually decreased by 1% per year, sowin a sarply declinin trend of labour productivy. Tis study analyses te trend in productivy in Brazilian and Mexican manufacturin industries between 1995 and 2009, a period in wic te two economies faced rowin international competion. Te analysis considers 14 sectors of manufacturin industry: food, beveraes and tobacco; textile and textile products; leater and footwear; wood and products of wood; paper and pulp; 2 coke and refined petroleum; cemical products, plastics and rubber; non-metallic mineral products; metallury and metal products; macinery and equipment; electrical and optical equipments; transport equipment; and oter industrial products. Te productivy trend is analysed in two ways: (i) usin te Leontief (1951) model to measure te consumption of intermediate oods used in production, and (ii) trou total factor productivy (tfp), wic takes account of production factor requirements. Te first measure of productivy defines te quanties of oods and services needed to produce one monetary un of a iven manufacture. Te analysis allows for comparisons of productivy trou time and space; and relative canes in productivies can be identified in te comparison between two countries over time. Noneteless, variations in production coefficients trou time do not necessarily imply an improvement or worsenin of tecnical and economic condions in te industrial sector in question. Amon oter tins, an industry s input expenses may rise because certain staes of production are outsourced. If tis step is taken to enance efficiency, te price of te oods may even fall, suestin a reduction in output value and an apparent loss of productivy. Noneteless, outsourcin saves on capal and labour in te final activy sector, involvin an increase in tfp. In tat case, a more detailed 2 Te paper and pulp sector includes production by te rapics and printin industry. A comparative analysis of productivy in brazilian and Mexican manufacturin industries Armênio de Souza Ranel and Fernando Garcia de Freas
cepal review 115 april 2015 183 analysis of productivy sould be complemented from te standpoint of te production factors used. Tat aspect is ily relevant in te comparison between Brazil and Mexico, since bot countries underwent trade liberalization wic, in eneral, increased te sare of imported oods and services in industry s intermediate consumption. Brazil also saw intensive outsourcin in manufacturin activies, owin to te risin costs of labour and social protection processes tat ad been under way since te early 1990s. Outsourcin dynamics were also impacted by te deree of economic interation, wic was very different in te two countries. Tis article is divided into tree sections apart from tis introduction. Section II compares te industrial productivy of te Brazilian and Mexican economies from te standpoint of te consumption of intermediate inputs, wereas section III analyses tfp. Lastly, section IV summarizes and comments on te results of te analysis and briefly evaluates te influence of economic liberalization on te trend of industrial productivy in te two countries. II Input productivy 1. Te concept of productivy in input-output analysis Te lerature on input-output analysis describes tree widely used metods to evaluate tecnical cane: (i) te direct comparison of tecnical coefficients; (ii) structural decomposion, and (iii) te rowscaler metod. 3 All tree tecniques are based on te Leontief model, and teir applications use national input-output tables as data sources. Te direct comparison of tecnical coefficients was suested by Leontief imself (1951) as a way to evaluate tecnical cane. Considerin te basic equation of te Leontief model, X = (I - A) -1 Y = BY, in wic X is te vector of production, Y is te vector of final demand and B is te Leontief matrix, defined as te inverse of te difference between te identy matrix (I) and te tecnical coefficients matrix (A), te metod suested by Leontief entails directly comparin te a ij of two A matrices, wic can differ in time or space. 4 Wen tis is applied to matrices of pysical coefficients, te metod adms only partial conclusions, because is impossible to areate quanties to identify te caracteristics of a sector, for example. Altou areation is possible in te case of monetary matrices, te metod as 3 Based on anoter study by Carter (1980), Feldman, McClain and Palmer (1987) proposed a metod for comparin matrices w incomplete data. Te study in question describes an adapted version of te oriinal ideas for square matrices, takin account of te direct and indirect effects on te matrices. 4 Te first case would evaluate tecnical cane, wereas te second would estimate te tecnoloical differences between two economies w different tecnoloies. sortcomins, because supports evaluation of cost trends only, wic could stem from tecnical canes or sifts in input prices, or bot. 5 Structural decomposion as also been widely used in evaluatin tecnical canes. 6 Tis metod consists of breakin down te sources of te variation in ross production value. Based on te production equation, X = BY, te total variation in ross production X can be spl into tree parts, as sown in te followin equation: DX = BDY+ YDB+ DBDY (1) Accordin to tat expression, differences in te value of production owin to canes in final demand can be estimated by settin te matrix of tecnical coefficients: X = B Y. Differences in output value resultin from canes in tecnical coefficients are obtained by settin te vector of final demand: X = Y B. W tis metod, tecnical cane is estimated by te difference in te tecnical coefficients between te two matrices wic, to obtain te same net output, use different amounts of intermediate inputs. Te reater te quanty, or value, of tose inputs, te lower productivy will be. It is also possible to identify te sectors of te economy wic, in te areate, record te larest canes between two points in time, 7 and, in turn, te coefficients responsible for te cane. 5 Tat areation reveals te production cost of products or sectors. 6 On tis point, see Lar and Dietzenbacer (2001). 7 Tis is possible only wen te matrix is expressed in monetary values. A comparative analysis of productivy in brazilian and Mexican manufacturin industries Armênio de Souza Ranel and Fernando Garcia de Freas
184 cepal review 115 april 2015 Structural decomposion can also be used to identify (approximately) te differences between two matrices in time and space. To specifically measure te impact of te tecnical canes, te vector of final demand of te economy must be set between two periods, allowin only te tecnical coefficients matrix to vary. Carter (1967) pioneered a study tat analysed te tecnoloical canes tat ave taken place in te Uned States economy, comparin te matrices of 1947 and 1958. Given te vector of final demand for 1962 and te coefficients of te inverse matrices of 1947 and 1958, te autor obtained te ross production vector for te sectors of te Uned States economy compatible w tat demand and eac period s Leontief matrices. Te difference between te two ross production vectors determines te variation in production needed to satisfy te same final demand vector in te two periods. A posive variation would imply a productivy loss, because te same demand would require a larer amount of expendure to produce te oods or services of te sector in question. In contrast, a neative variation would mean a reduction in expendure and a consequent increase in productivy. All of te sectors of te economy can be areated to determine wic of te two matrices is more productive. Altou tis metod allows for direct comparisons to be made between sectors of two tecnoloical matrices, comparin two economies poses a number of problems. Tis is because te areation depends on te composion of te ross production vector wic, in turn, depends on te composion of final demand, a variable tat is exoenous to te system. Differences in te composion of demand can produce different results. Wereas te sare of low-productivy sectors in te economy as a wole tends to decline, te sare of i-productivy sectors tends to increase. As a result, attributin te same sare to te productive sectors of te two matrices could lead to distortions in te analysis. Carter (1967), for example, imposed te 1962 composion of final demand on te years 1947 and 1958. 8 Feldman, McClain and Palmer (1987) proposed a rowscaler metodoloy. To explain, te autors start w two different tecnoloical matrices (eac one associated w a point in time), and a sinle vector of final demand. Te values of ross output needed to obtain te final demand vector are iven by: -1 X t = A t X t + Y = _ I A t i Y = B t Y, t = 12, (2) 8 If Carter (1967) ad used a final demand vector w different composion, te result would probably ave been different. Te diaonalization of te production vectors obtained in eac period enerates te diaonal matrices Xt t, wic ave te values of te production vectors in te leadin diaonal, and zeros in te oter cells. Multiplyin te diaonalized production vector obtained in te first period by tat obtained in te second bot based on te same demand but w different tecnoloies ives expression (3): C = Xt -1 2 1 7Xt A (3) were Γ Is te matrix 9 formed by te γ ij elements. If c ii 2 2 1, sector i of matrix 2 is less productive tan te same sector of matrix 1; if c ii 2 1 1, sector i of matrix 2 is more productive tan te correspondin sector of 2 matrix 1. Lastly, if c ii = 1, te two matrices can be said to ave te same productivy. In te specific case were te final demand vector is unary, is possible to directly compare te direct and indirect coefficients of te inverse Leontief matrix for te two countries. Te sum of eac of te rows of te Leontief matrix would indicate te direct and indirect quanty of oods and services needed to obtain one un of ood i to satisfy final demand. 10 Te foreoin metod takes account of te production of all oods needed to satisfy a unary demand vector, wose elements contain one un of demand for eac ood and service in te economy, in oter words a vector w un values in eac row. Wen te aim is to investiate te quanty of oods needed to satisfy te demand for one un of a iven ood i, te metod consists of summin te rows of te Leontief matrix correspondin to tat ood. Tat process areates te direct and indirect quanties of oods and services needed to produce one un of te ood or service bein analysed. In addion to te tree metods described above, input-output analysis also developed a specific metodoloy to evaluate te trend of tfp, in te tradion 9 In tat matrix, te elements outside te leadin diaonal are zero by construction. 10 Tis procedure makes possible to compare te same sectors of different matrices to ascertain weter a specific sector is more or less productive tan te equivalent sector in te oter economy. It is impossible to determine weter one economy is more or less productive tan te oter. Only te extreme case were all sectors of one of te matrices are more productive tan te respective sectors of te oter matrix, would be possible to state, unequivocally, tat one of te matrices is more productive tan te oter. To compare areate production, would be necessary to set a vector of final demand or production, as proposed by Carter (1967). A comparative analysis of productivy in brazilian and Mexican manufacturin industries Armênio de Souza Ranel and Fernando Garcia de Freas
cepal review 115 april 2015 185 of economic rowt teories. 11 Noneteless, te analysis cannot be applied to te international input-output matrix database, because tis does not contain information on factor endowments (capal, labour and land) for all te economies. To determine te trend of tfp in te manufacturin sectors of te two countries, te followin subsection uses anoter teoretical approac based on statistical metods. 2. Databases and metodoloy Te databases used in te analysis are te wiod (2012) lobal input-output tables. Te tables were constructed as a result of a ue task to make national input-output tables compatible w one anoter, undertaken by a roup of researc instutes around te world, coordinated by te Universy of Groninen. 12 Te project, financed by te European Commission and publised on te Internet in April 2012, will make a major contribution to deepenin understandin of te world economy. 13 Te wiod as data on national input-output tables for te years between 1995 and 2009, toeter w estimations of lobal input-output tables tat sow international flows of oods and services. Te data cover a total of 40 countries, and estimations for te rest of te world te reional roupin created to reconcile national forein trade flows. In addion to te national and international matrices, numerous oter variables are provided per country, suc as factor endowments, price indices, and te functional distribution of income. Tis article uses te lobal input-output tables for 1995-2009 to calculate te productivy vector. Tese matrices abide by te oriinal Leontief formulation, X = (I A) -1 Y, were: R 1V R 1V R 11 SX W SY W SA SW SW S S 2W S 2W S 21 SX W SY W SA X= SW, Y= SW and A= S S W S W S S W S W S SW SW S SX W S 1 Y W SA T X T X T A 22 A 2 A k k k k 12 j 1kV A W W 2kW A W W(4) W W W kk A W X 11 On tis topic, see Miller and Blair (2009). 12 Te edor in care is Marcel Timmer of te Universy of Groninen. 13 [Online] www.wiod.or/database. In expression (4), X 1, X 2,..., X k are (35 x 1) vectors of national production, eac of wic contains te output values of te 35 sectors of economic activy covered by te matrix. Te vector Y denotes final demand and as te same interpretation. Te matrix is formed by 1,681 matrices of tecnical coefficients (of dimension 35 x 35) wic identify te oriin (country and sector) and destination (country and sector) of intermediate consumption. 14 Matrix A is calculated by dividin te intermediate consumption matrix, of te same dimension as matrix A, by vector X. Tose data are used to obtain te lobal Leontief matrix for te years 1995 and 2009, wic serves as a basis for calculatin te production needed to satisfy one un of final demand for a iven ood or service in a specific country. Expression 5 contains te definions of tat matrix, in wic N and M indicate te countries, and B NM te sub-matrix of tecnical coefficients of tose two countries. In tis system, wen N is equal to M, te matrix B NM desinates te domestic coefficients of economy N, in oter words te quanties of oods produced in economy N tat are needed to produce one un of te ood in tat economy. Wen N is different from M, te matrix B NM denotes te external coefficients of economy N: te quanties of oods and services produced in te rest of te world tat are needed to produce one un of te ood in economy N. R 11 12 SB B S S 21 22 SB B B = S S S S S k1 k2 B B T and R Sb11 b Sb b NM 21 B = S S Sbi 1 bi T 12 22 2 j j V b1jw b W 2j W W b W ij X NM 1kV B W W 2kW B W W W W W kk B W X (5) As noted above, te production needed to satisfy one un of demand for a ood in a iven country is calculated by addin te values of te columns of te Leontief matrix. Tis can be done directly w respect 14 Te number of A kk matrices stems from te number of reions in te database (41). A comparative analysis of productivy in brazilian and Mexican manufacturin industries Armênio de Souza Ranel and Fernando Garcia de Freas
186 cepal review 115 april 2015 to all economic sectors and countries of interest. 15 Te resultin sum can be broken down into two elements representin domestic and external requirements. Te domestic requirements are calculated as te sum of te b ij values in te B NN matrix, and te external requirements are calculated as te sum of bij values of te B NM matrices, N M. Te ratio between te quanties needed in te two countries enerates te matrix Γ, wose meanin and interpretation are identical to tose presented in subsection 2 of section II. In te followin analysis, two oter matrices are calculated w tose caracteristics: Γ d, wic indicates te quanties of domestically produced oods and services tat are needed to produce i oods in te two economies; and Γ e, calculated as te ratio between te quanties of imported oods and services used in te production of i oods in te two economies. Tis article only presents te values for te Brazilian and Mexican economies in te 14 activy sectors of manufacturin industry mentioned in s introduction. Te values of te lobal input-output tables are expressed in millions of dollars at current prices eac year. As te productivy indicators are expressed as monetary uns of production of te oods, te inter-temporal comparisons contain variations in bot quanties and relative prices, wic restrict analysis possibilies. It sould be remembered tat te analyses do not consider differences in te purcasin power of currencies. In 2009, one dollar in Brazil was equivalent in purcasin power to US$ 1.32 at Uned States prices, wereas in Mexico, one dollar was wort US$ 2.19 in purcasin-power terms. Tat meant tat a dollar of demand for a iven ood in Brazil was equivalent to a different amount of tat ood in Mexico. Noneteless, relative production studies do not need adjustments for purcasin power pary, since tey express relations between production values at a iven place and time. 3. Results Altou Brazil s economy is muc larer tan Mexico s, teir manufacturin industries display very similar structures. In 2009, te ross value of industrial output amounted to US$ 940,559 million in Brazil and US$ 470,853 million in Mexico (see table 1). In te Brazilian economy, 73% of ross production value was accounted for by five industries: food, beveraes and tobacco; coke and refined petroleum; cemicals; 15 Tis is equivalent to calculatin te value of X needed to satisfy a un vector Y. metallury and metal products; electrical and optical equipment; and transport equipment. Te equivalent sare was even reater in te case of Mexico, at 80.5%. Te structural cane coefficient, wic ad a value of 0.89 in 2009, illustrates te similary between te two industrial structures and between te two economies more clearly. 16 It is also wort ilitin te importance of te food, beveraes and tobacco industry in te two countries industrial structure. Tis sector accounted for 19.8% of Brazil s ross industrial production value in 2009, and 24.6% in te case of Mexico. Te reatest difference between te structures of te two countries corresponds to te electrical and optical equipment sector, wic represented almost 15% of Mexican industry in 2009 compared to 6.6% in te case of Brazil. Table 2 reports productivy indicators for 1995, in terms of te quanties of oods (domestic and imported) needed to produce one monetary un of te oods of eac industry in eac country. Te last tree columns sow te ratios between tose amounts in te two countries. Takin te data for Brazil as an example, in 1995 te food, beveraes and tobacco industry required US$ 2.27 for eac dollar of output, representin US$ 0.12 in imported oods and services plus US$ 2.14 of oods and services produced domestically. In Mexico, to produce one dollar of food, beveraes and tobacco required US$ 2.22 of production in all sectors of te economy US$ 0.32 of imported oods and services, and US$ 1.90 of domestic production. Te ratio between te two coefficients of production in te food, beveraes and tobacco sector was 1.02 in 1995, wic sows tat te Brazilian industry was slitly less productive tan s Mexican counterpart. Table 2 also reports te values of te leadin diaonal of te Γ d matrix, wic relates te quanties of domestically produced oods and services tat are needed for te production of food and beveraes in te two economies. In te case of te food, beveraes and tobacco industry, te ratio was 1.13. Tis means tat te Brazilian food and beveraes industry required more monetary uns of domestic production tan s Mexican equivalent, and tat te productive cain in Mexico also required more imports. 17 16 Coefficient of correlation between te distribution (percentae) of ross production value between te two economies. 17 Tose calculations do not include imports of manufactured food and beveraes by te two countries, but only te imports of raw materials (oods and services) needed to produce one monetary un of te oods in question. A comparative analysis of productivy in brazilian and Mexican manufacturin industries Armênio de Souza Ranel and Fernando Garcia de Freas
cepal review 115 april 2015 187 Table 1 Brazil and Mexico: ross industrial production value, 2009 (US$ million) Sector Brazil Percentae Mexico Percentae Total Percentae Food, beveraes and tobacco 186 480 19.8 115 636 24.6 302 136 21.4 Textiles and textile products 42 162 4.5 14 380 3.1 56 546 4.0 Leater and footwear 14 037 1.5 4 511 1.0 18 549 1.3 Wood and products of wood 11 825 1.3 3 588 0.8 15 414 1.1 Paper and pulp 44 507 4.7 17 294 3.7 61 806 4.4 Coker and oil refinin 92 996 9.9 39 495 8.4 132 502 9.4 Cemical products 111 678 11.9 45 287 9.6 156 977 11.1 Plastics and rubber 32 121 3.4 12 030 2.6 44 155 3.1 Nonmetallic mineral products 26 381 2.8 16 156 3.4 42 540 3.0 Metallury and metal products 113 573 12.1 46 327 9.8 159 912 11.3 Macinery and equipment 59 588 6.3 9 672 2.1 69 266 4.9 Electrical and optical equipment 62 065 6.6 70 421 15.0 132 493 9.4 Transport equipment 119 805 12.7 61 750 13.1 181 568 12.9 Oter industrial products 23 341 2.5 14 306 3.0 37 650 2.7 Total 940 559 100.0 470 853 100.0 1 411 512 100.0 Source: prepared by te autors, on te basis of information from te World Input-Output Database (wiod). Te data of table 2 sow tat, in 1995, 10 sements of Brazilian industry were more productive tan teir counterparts in Mexico, namely: textiles and textile products; wood and products of wood; paper and pulp; cemical products; plastics and rubber; metallury and metallic products; macinery and equipment; electrical and optical equipment; transport equipment; and oter industrial products. Only in te leater and footwear, non-metallic mineral products, and coke and refined petroleum sectors did te productivy of Mexican industry reatly exceed tat of Brazil. In te specific case of coke and refined petroleum, te productivy difference was larely due to te reater need for imports in Brazilian industry, wic was not yet self-sufficient in oil production. Te data of fiure 3 sow a very different suation in 2009, because Mexico s industry ad overtaken Brazil s in productivy terms. All industrial sements, Table 2 Brazil and Mexico: input requirements and relative productivy, 1995 Industrial sectors Brazil Production needed Mexico Γ Total Domestic External Total Domestic External Total Domestic External Food, beveraes and tobacco 2.2676 2.1436 0.1241 2.2244 1.8975 0.3269 1.0194 1.1297 0.3795 Textiles and textile products 1.9997 1.8188 0.1809 2.3634 1.8648 0.4987 0.8461 0.9753 0.3628 Leater and footwear 2.4597 2.2514 0.2084 2.3231 1.9371 0.3860 1.0588 1.1622 0.5398 Wood and products of wood 1.8489 1.7591 0.0898 2.1605 1.8841 0.2764 0.8558 0.9337 0.3249 Paper and pulp 2.0535 1.8796 0.1739 2.0616 1.6402 0.4214 0.9961 1.1459 0.4127 Coker and oil refinin 2.5341 2.1945 0.3397 2.1276 1.9858 0.1418 1.1911 1.1051 2.3953 Cemical products 2.1038 1.8931 0.2107 2.1552 1.8632 0.2920 0.9762 1.0161 0.7215 Plastics and rubber 2.2005 1.9492 0.2513 2.3242 1.7562 0.5679 0.9468 1.1099 0.4425 Nonmetallic mineral products 1.9870 1.8371 0.1499 1.8275 1.5988 0.2287 1.0872 1.1490 0.6554 Metallury and metal products 2.1255 1.9193 0.2062 2.3187 1.7878 0.5309 0.9167 1.0736 0.3884 Macinery and equipment 2.1956 1.9914 0.2043 2.2692 1.5589 0.7104 0.9676 1.2775 0.2875 Electrical and optical equipment 2.3319 2.0319 0.3000 2.7552 1.4770 1.2782 0.8464 1.3757 0.2347 Transport equipment 2.4428 2.1795 0.2633 2.4756 1.6337 0.8419 0.9867 1.3341 0.3127 Oter industrial products 1.9878 1.8411 0.1467 2.3174 1.6576 0.6598 0.8578 1.1107 0.2224 Source: prepared by te autors, on te basis of information from te World Input-Output Database (wiod). A comparative analysis of productivy in brazilian and Mexican manufacturin industries Armênio de Souza Ranel and Fernando Garcia de Freas
188 cepal review 115 april 2015 apart from te textile products, electrical and optical equipment, and oter industrial products industries were less productive in Brazil tan in Mexico. Tose tree sectors were already more productive in Brazil in 1995, and te advantaes w respect to Mexico ad diminised in two of tem by 2009. In contrast, te four sectors of Mexican industry tat were more productive in 1995 ad actually increased teir advantae by 2009. In tis suation, is natural to ask ow a team tat won a matc 10-4 can lose a second one 3-11? Table 4, wic reports te rates of cane of te indicators sown in te previous tables between 1995 and 2009, answers tat question. Comparin te data for te two reference years, tere are sinificant canes in te oods and services requirements in te two countries; and Brazilian industry recorded reater increases in all sectors, except electrical equipment. In Brazil, requirements increased in all industrial sectors, except for te leater and footwear industry. Tis indicates a loss of productivy, wic could represent bot a pysical decline and an adverse trend in relative prices. In te Mexican case, tere were considerable productivy ains in eit of te 14 manufacturin industries between 1995 and 2009. Te productivy loss in te oter sectors was less tan in Brazilian industry, except in te case of electrical equipment. Anoter important caracteristic of Mexican industry is tat te rowt in imported oods and service requirements was offset by a reduction in requirements for domestic oods and services. Tus, compared to te Brazilian case, te productivy of Mexican industry evolved by replacin domestic raw materials w imports, in oter words, tat trade liberalization enerated larer productivy ains in Mexico tan in Brazil. 18 18 Te only exception to tat rule was te coke and refined petroleum industry. Table 3 Brazil and Mexico: input requirements and relative productivy, 2009 Industrial sectors Brazil Production needed Mexico Γ Total Domestic External Total Domestic External Total Domestic External Food, beveraes and tobacco 2.5257 2.3525 0.1732 2.1461 1.7900 0.3560 1.1769 1.3142 0.4864 Textiles and textile products 2.1539 1.9064 0.2475 2.2520 1.6807 0.5713 0.9564 1.1343 0.4332 Leater and footwear 2.3820 2.1767 0.2054 2.1947 1.7799 0.4148 1.0854 1.2229 0.4952 Wood and products of wood 2.1030 1.9517 0.1514 2.0346 1.7229 0.3117 1.0336 1.1328 0.4855 Paper and pulp 2.1271 1.9196 0.2075 2.0509 1.6050 0.4460 1.0371 1.1960 0.4653 Coker and oil refinin 2.7676 2.3908 0.3768 2.2069 1.9969 0.2100 1.2541 1.1973 1.7941 Cemical products 2.5001 2.1684 0.3317 2.1988 1.7836 0.4152 1.1370 1.2157 0.7990 Plastics and rubber 2.4443 2.0914 0.3530 2.3837 1.6912 0.6925 1.0254 1.2366 0.5097 Nonmetallic mineral products 2.2249 2.0207 0.2042 1.8193 1.5656 0.2537 1.2230 1.2907 0.8049 Metallury and metal products 2.3269 2.0570 0.2700 2.2955 1.6944 0.6012 1.0137 1.2140 0.4491 Macinery and equipment 2.4189 2.1234 0.2955 2.3393 1.5334 0.8059 1.0341 1.3848 0.3667 Electrical and optical equipment 2.5327 2.0422 0.4905 3.0350 1.4594 1.5756 0.8345 1.3993 0.3113 Transport equipment 2.7291 2.3044 0.4246 2.4459 1.5720 0.8739 1.1158 1.4659 0.4859 Oter industrial products 2.1408 1.9348 0.2060 2.3324 1.5828 0.7496 0.9178 1.2224 0.2748 Source: prepared by te autors, on te basis of information from te World Input-Output Database (wiod). A comparative analysis of productivy in brazilian and Mexican manufacturin industries Armênio de Souza Ranel and Fernando Garcia de Freas
cepal review 115 april 2015 189 Table 4 Brazil and Mexico: variation in input requirements and relative productivy, from 1995 to 2009 (Percentaes) Industrial sectors Brasil Production needed México Γ Total Domestic External Total Domestic External Total Domestic External Food, beveraes and tobacco 11.4 9.7 39.6-3.5-5.7 8.9 15.4 16.3 28.2 Textiles and textile products 7.7 4.8 36.8-4.7-9.9 14.6 13.0 16.3 19.4 Leater and footwear -3.2-3.3-1.4-5.5-8.1 7.5 2.5 5.2-8.3 Wood and products of wood 13.7 10.9 68.5-5.8-8.6 12.8 20.8 21.3 49.4 Paper and pulp 3.6 2.1 19.3-0.5-2.1 5.8 4.1 4.4 12.7 Coker and oil refinin 9.2 8.9 10.9 3.7 0.6 48.1 5.3 8.3-25.1 Cemical products 18.8 14.5 57.5 2.0-4.3 42.2 16.5 19.7 10.7 Plastics and rubber 11.1 7.3 40.4 2.6-3.7 21.9 8.3 11.4 15.2 Nonmetallic mineral products 12.0 10.0 36.2-0.5-2.1 10.9 12.5 12.3 22.8 Metallury and metal products 9.5 7.2 30.9-1.0-5.2 13.2 10.6 13.1 15.6 Macinery and equipment 10.2 6.6 44.7 3.1-1.6 13.4 6.9 8.4 27.5 Electrical and optical equipment 8.6 0.5 63.5 10.2-1.2 23.3-1.4 1.7 32.7 Transport equipment 11.7 5.7 61.3-1.2-3.8 3.8 13.1 9.9 55.4 Oter industrial products 7.7 5.1 40.4 0.6-4.5 13.6 7.0 10.1 23.5 Source: prepared by te autors, on te basis of information from te World Input-Output Database (wiod). III Factor productivy As is impossible to apply input-output matrix analysis owin to te lack of data on factor endowments (capal, labour and land) in te set of countries formin te area referred to as rest of te world in te wiod, factor productivy in te 14 industrial sectors of Brazil and Mexico was evaluated usin a different approac. In tis case, tfp was calculated on te basis of te Solow residual (Solow, 1957). To improve te analysis of productivy trends, separatin te effects of demand and supply crises on tat indicator from loner-term trends (suc as tecnoloical proress and economies of scale), a complementary statistical approac was used. Tis firstly involves ftin a production function and ten usin te estimated coefficients to calculate productivy by means of a decomposion. Under tis approac, te productivy trend is te portion of dp rowt tat is explained neer by factor accumulation a concept present in te Solow (1957) approac nor by specific random penomena. 19 1. Production frontier and decomposion of productivy Tis study adopted te stocastic-frontier econometric approac to f te production function. Tis approac 19 It is not necessary to estimate a production function to calculate tfp. Te calculation can be based on statistics of te trend of dp, factor endowments and te factor sares in te functional distribution of income. Noneteless, a strictly accountin approac accentuates te effects of supply and demand crises in te measurement of productivy trends. Econometric approaces, on te oter and, make possible to remove random penomena from te variations in dp and factor endowments, and, dependin on te tecnique, measurement errors. In eneral, tose approaces produce more stable tfp estimations w more plausible economic interpretations. A comparative analysis of productivy in brazilian and Mexican manufacturin industries Armênio de Souza Ranel and Fernando Garcia de Freas
190 cepal review 115 april 2015 as been, widely applied in microeconomic studies and was used w satisfactory results in recent studies for te international comparisons of factor productivy at more areate levels. On tis point, see Kneller and Stevens (2003); Kumbakar and Wan (2005); Garcia, Souza and Pires (2008), and Pires and Garcia (2012). Te first advantae of te approac is tat te productivy difference between two economies is not restricted to tecnoloical differences. Te stocastic production frontier adms te possibily of inefficiency in production and, terefore, tat tere may be productivy differences between two economies tat operate at te same tecnoloical level. Anoter advantae is tat, wen panel data are used, te stocastic frontier produces better estimates ten ordinary least squares (ols) in te absence of eteroeney controls. 20 Tis is because is based on an error component model tat makes possible to separate random penomena from tose tat can be attributed to omted factors, suc as te output ap caused by labour unemployment. Expression (6) defines te stocastic production frontier as a production function fted trou a teoretical measure of tecnical inefficiency. Y = F _ Bt, K, H, Li$ exp_ uii, i = 1, 2,..., N and t = 1,..., T (6) Tis expression uses te followin definions: Y is industrial value added in country i at time t; F is te production function; K, H and L are te quanties of capal, skilled labour and unskilled labour used by te industry of country i at time t; B t is te level of productivy reflectin optimal practice at time t, and u 0 is te measure of tecnical inefficiency of te industry of country i at a time t. Based on te stocastic production frontier, te Bauer-Kumbakar decomposion of te trend of tfp is performed (see Kumbakar, Denny and Fuss, 2000), to identify four sources of productivy variation: tecnical proress, variation in tecnical inefficiency, variation 20 As sown in Garcia, Souza and Pires (2008), areate production functions wic control te eteroeney produce estimations wout economic meanin; for example, te African economies would sow te iest rate of tecnoloical proress wile te industrial economies would display reression or stanation. Tis appens because tere is a very close linear link between capal, tecnoloy and te quanty of labour used, because te tecnoloies are embedded in te capal. of allocative inefficiency, and economies of scale. Te decomposion also makes possible to interpret te trend of tfp more precisely and to identify different patterns. For example, altou two economies may display te same tfp rowt rate, in one case te increase may stem from tecnoloical proress and in te oter from economies of scale, wic are very different economic processes. In matematical terms, te decomposion of te productivy trend under te production-frontier model is obtained by differentiatin tat frontier w respect to time. Followin numerous alebraic manipulations, te time differential of te production frontier ives equation (7), wic expresses te rate of rowt of industrial value added in a iven country i at a time t, as te sum, weited by te respective elasticies (e), of te rates of: (i) variation in optimal practice, also known as te rate of tecnoloical proress; (ii) factor accumulation (capal, skilled and unskilled labour), and (iii) variation in tecnical inefficiency. Yo Bo Ko Ho Lo = f u Y B $ + f B K$ + f K H$ + f H L $ o (7) L Assumin Hicks-neutral tecnoloy wic means tat ε B = 1, and usin te tfp definion establised by equation (8) te Solow residual in wic S j is te sare of productive factor j in te functional distribution of income, is possible to find a new decomposion for te variation of tfp. Ao Yo Ko Ho Lo = S S S A Y K$ K H$ H L $ L Combinin (7) and (8), ives: Ao Bo Ko Lo = + S S A B _ fk Ki$ + K _ fl Li$ L Ho + _ fh SHi $ uo H (8) (9) A transformation can be applied to simplify te foreoin expression (9) and isolate te components of te rate of cane of tfp. Definin: / f j RTS = fj and m =,,,, RTS j K HL j j = A comparative analysis of productivy in brazilian and Mexican manufacturin industries Armênio de Souza Ranel and Fernando Garcia de Freas
cepal review 115 april 2015 191 in wic rts denotes returns to scale, ives equation (10): Ao Bo Ko Ho Lo = uo RTS 1 A B + _ i $ > mk $ + m K H $ + m H L $ H L (10) Ko Ho Lo + > _ mk SKi$ + m S S K _ H Hi$ + m H _ L Li $ H L wic states tat te rate of cane of tfp can be broken down into four components: (i) tecnical proress measured by B o B ; (ii) cane in tecnical efficiency approximated by uo ; (iii) cane in productivy owin to te effect of a cane in te scale of production, calculated by Ko Ho Lo _ RTS 1i $ > mk $ + m K H $ + m H L $ H, L (iv) cane in allocative efficiency, measured by Ko Ho Lo > _ mk SKi$ + m S S K _ H Hi$ + m H _ L Li $ H L Under constant returns to scale, rts = 1, so te tird component of te productivy cane is cancelled out; but if rts differs from 1, part of te variation in productivy is explained by te cane in te scale of production. Moreover, if te ratios between te elasticies and rts (l j ) are equivalent to te respective factor sares in te functional distribution of income (S j ), ten te industry is efficient in terms of factor allocation. In tat case, by definion, tere are no productivy canes attributed to canes in te allocation of factors. Lastly, in tis model, tecnical proress accounts for at least as muc of te variation in productivy. Only wen tere are no tecnical or allocative inefficiencies, or increasin or decreasin returns to scale, is te measure of te variation of productivy, Ao A, identical to tecnical proress, Bo B. Tis approac tus covers a larer number of possible suations, wout very arbrary restrictions on te sape of te production function and s properties. 2. Databases and econometric model Te data used in te analysis also come from te wiod and relate to te 14 manufacturin industries analysed above. For eac industrial sector, a stocastic frontier is estimated based on te data from 40 countries w reard to value added (Y ), capal endowment (K ), ours of skilled labour employed (H ) and ours of unskilled labour employed (L ) between 1995 and 2009. Hours of skilled labour employed are equivalent to te sum of te number of ours worked by medium- and i-skilled workers. Te monetary values are expressed in constant 1995 dollars. 21 As te analyses were conducted at te industrial-sector level, te data were not adjusted for purcasin power pary, as is more frequent in areate macroeconomic analyses. Te econometric model estimated is a translo function of te value added of te tree factors of production and time (t), wic captures te trend of te frontier. Te function in question, described in equation (11), as 14 explanatory variables: te levels of factors of production and time (K, H, L and t), te squares of te factors of production and time (K 2, H 2, L 2 and t 2 ) and te interactions between tem (K.H, K.L, K.t, H.L, H.t and L.t). Te variables u and v are te model s error components: te first of tese measures tecnical inefficiency and as a distribution u ~ i.i.d N + `nv, u 2 j; and te second is te random error w distribution v ~ i.i.d N`0, vv 2 j. Te values of all variables (except time) are expressed in natural loarms and are deviations from te mean of eac series (includin time), suc tat te estimated coefficients of eac reression are fted to te sample mean. ln y = b0 + bt. t+ bkln K + blln L+ bhln H 2 2 + 1 2$ btt. t + 1 2$ bkk _ ln Ki 2 2 + 1 2$ bll _ ln Li + 1 2$ bhh _ ln Hi + bkt 9_ ln Ki$ tc+ bkl 9_ ln Ki$ _ ln LiC + bkh 9_ ln Ki$ _ ln HiC+ blt 9_ ln Li $ tc + blh 9_ ln Li$ _ ln HiC+ bht 9_ ln Hi $ tc + v u (11) As proposed by Garcia, Souza and Pires (2008), no coefficients are included to control for eteroeney between countries. Given te i correlation tat exists between te dummy variables and te explanatory variables, tat procedure enerally distorts te estimates of tecnical efficiency and tecnoloical proress. It is 21 Altou te total number of observations is 585, owin to te lack of data on te capal endowment in some countries, te number of effective observations in te panels varies between 570 and 547. A comparative analysis of productivy in brazilian and Mexican manufacturin industries Armênio de Souza Ranel and Fernando Garcia de Freas
192 cepal review 115 april 2015 terefore assumed tat any eteroeney in te industrial sectors of te sample countries can be captured trou te model s explanatory variables and te tecnical inefficiency component. Te trend of tfp is estimated usin equation (8). As tere is no information on labour remuneration by skill level in te wiod, te expression was simplified to encompass te total variation in ours worked, wout prejudice to te concepts defined in te foreoin section. Tecnoloical proress and te elasticies of value added w respect to eac factor of production are iven by equations (12) and (13). By construction, te elasticies and tecnical proress of a iven activy sector vary trou time and across countries. Te variation in allocative efficiency was obtained as a residual by definion, tat measure is te variation in tfp, avin discounted tecnical proress, tecnical efficiency and economies of scale. Bo TC = = b B t+ b $ t + b $ K + b $ H + b $ L tk th tl f = b + b $ K + b $ H + b $ L + b $ t, j j jk jh jl jt j = K, HL, tt (12) (13) Table 5 reports on te estimations of te coefficients of te production frontiers of te 14 industrial sectors for te 40 countries in te sample between 1995 and 2009. In nine of te 14 sectors te variance of te error term u is sinificantly different from zero, wic indicates productive inefficiency. In te oter sectors, inefficiency is relatively minor, and random deviations from te production frontier predominate. As te variance of u tends to zero in te model of te macinery and equipment sector, was estimated usin ols. Most of te coefficients are sinificantly different from zero at te 10% sinificance level in all models, wic sows tat te translo model is appropriate as a eneric specification of te frontiers. Moreover, te presence of non-sinificant coefficients is foreseen in tis type of analysis, since te number of observations (maximum 570) is relatively small for te set of parameters to be estimated (14). 3. Results Based on te foreoin estimations, te mean elasticies of te factors of production were firstly calculated in eac of te 14 industrial sectors of Brazil and Mexico, alon w te averae rate of tecnoloical proress. Tose data were aumented by te estimations of tecnical efficiency to evaluate te trend of tfp and s components in te two countries between 1995 and 2009. Table 6 sows te estimations for eac of te industries in Mexico and Brazil. As can be seen, te patterns of capal accumulation and te trend of productivy differ reatly between te two countries. In nearly all industrial sectors, capal accumulation rates are ier in Brazil tan in Mexico. A similar pattern can be seen in te use of skilled labour, as employment rowt is ier in Brazil. Tis trend is partly offset by a larer reduction in unskilled employment in Brazil tan in Mexico, wic indicates a more intensive rate of substution of labour by capal and uman capal in Brazil, probably reflectin te sarp rise in labour costs in tat country. Accordin to wiod data, te averae value of real waes 22 in Brazil rew by 3.1% per year between 1995 and 2009, compared to a reduction of 0.9% per year in Mexico. Te counterpart of te slower pace of factor accumulation in Mexico was more viorous rowt of tfp. Mexico s industries enerally recorded ier tfp rowt rates, except for te textile products, cemicals and macinery and equipment industries. In terms of te components of tfp, te suation is que varied. In Mexico, tecnoloical proress between 1995 and 2009 was posive in all industrial sectors except for five. In Brazil, owever, eit of te 14 sectors posted neative rates of tecnoloical proress, probably owin to te recomposion of production win eac sector, involvin te retreat of ier value added product lines and specialization in products of lower tecnoloical content. Te reduction in manufacturin industry value added per our worked, mentioned in te introduction to tis article, corroborates tat idea. Also important is te influence of te eneral economic suation in 2009, te last year of te comparison, because te lobal recession te prices of several industrial products, w effects on industrial value added. Te trend of tecnical efficiency is also worse in Brazil tan in Mexico in 10 of te 14 sectors analysed. Te suation is even more serious in terms of economies of scale: as sown in table 6, Brazil las beind Mexico in 11 out of te 14 industrial sectors. Tis is probably affected by Mexico s trade interation w te Uned States and Canada, wic considerably expands te scale of businesses in te country. In te case of allocative 22 Variation over and above te inflation rate (consumer prices). A comparative analysis of productivy in brazilian and Mexican manufacturin industries Armênio de Souza Ranel and Fernando Garcia de Freas
cepal review 115 april 2015 193 Table 5 Brazil and Mexico: estimations of stocastic frontiers, 14 industrial sectors, coefficients and p-value Industrial sectors Translo coefficients Statistics K L H t KK KL KH Kt LL LH Lt HH Ht tt Lns 2 v lns 2 u lo of MV Food, beveraes and tobacco 1.0165-0.0843 0.1012-0.0168-0.1821-0.1180 0.2036 0.0071 0.0170 0.0388-0.0005-0.1652 0.0001-0.0002-1.6480-9.1544 p-value 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.1640 0.5950 0.1360 0.9100 0.0000 0.9820 0.9420 0.0000 0.8930-339.172 Textiles and textile products 0.9299-0.0189 0.0706-0.0136-0.2966 0.0553 0.1797 0.0034-0.0003-0.0236 0.0018-0.1318 0.0012-0.0038-2.8526-2.4878 p-value 0.0000 0.2870 0.0000 0.0000 0.0000 0.0000 0.0000 0.2950 0.9900 0.2480 0.5390 0.0000 0.7170 0.0230 0.0000 0.0000-113.743 Leater and footwear 0.7813 0.0703 0.0428-0.0116-0.1854-0.0319 0.1850-0.0082-0.0536 0.1015 0.0012-0.2908 0.0091-0.0016-2.4084-0.8007 p-value 0.0000 0.0090 0.0970 0.0550 0.0000 0.2490 0.0000 0.1230 0.1760 0.0020 0.8040 0.0000 0.0910 0.5320 0.0000 0.0000-393.624 Wood and products of wood 0.9257-0.1046 0.1795-0.0207-0.2431-0.0250 0.2325 0.0102 0.0661-0.0168-0.0101-0.1817-0.0009-0.0031-2.8903-0.8582 p-value 0.0000 0.0000 0.0000 0.0000 0.0000 0.1960 0.0000 0.0400 0.0260 0.4910 0.0060 0.0000 0.8590 0.1670 0.0000 0.0000-341.101 Paper and pulp 0.8537-0.0208 0.1417-0.0100-0.3205-0.0162 0.2381-0.0053-0.0212 0.0206 0.0040-0.2103 0.0063-0.0015-2.0437-1.2604 p-value 0.0000 0.3050 0.0000 0.0470 0.0000 0.4900 0.0000 0.3180 0.3890 0.4110 0.3400 0.0000 0.2170 0.5530 0.0000 0.0000-388.657 Coker and oil refinin 0.6201 0.3131 0.1780 0.0010 0.0758-0.1653 0.1291-0.0069-0.1278 0.1182-0.0083-0.1870 0.0129 0.0012-1.4531 0.0953 p-value 0.0000 0.0000 0.0000 0.9070 0.0000 0.0000 0.0000 0.1810 0.0000 0.0000 0.1680 0.0000 0.0570 0.7670 0.0000 0.5160-638.375 Cemical products 1.0143-0.0226-0.0082-0.0001-0.0589-0.0747 0.1197-0.0103-0.0251 0.1030 0.0027-0.2276 0.0102 0.0010-2.2197-2.2384 p-value 0.0000 0.1160 0.6550 0.9880 0.0170 0.0000 0.0000 0.0070 0.1680 0.0000 0.4060 0.0000 0.0100 0.6140 0.0000 0.0000-262.409 Plastics and rubber 1.0135-0.1415 0.1173-0.0069-0.0294-0.1010 0.1294-0.0124 0.1343-0.0221 0.0016-0.1062 0.0115-0.0028-3.7165-1.1710 p-value 0.0000 0.0000 0.0000 0.0520 0.4470 0.0000 0.0000 0.0010 0.0000 0.2080 0.5580 0.0000 0.0010 0.0980 0.0000 0.0000-211.317 Nonmetallic mineral products 1.0280-0.0649 0.0742-0.0065-0.2005 0.0262 0.1353-0.0031-0.0206-0.0236 0.0009-0.0688 0.0022-0.0037-2.4167-9.5583 p-value 0.0000 0.0000 0.0000 0.0430 0.0000 0.1130 0.0000 0.4250 0.2990 0.2030 0.7220 0.0140 0.5460 0.0190 0.0000 0.8610-120.106 Metallury and metal products 0.8969-0.1358 0.1963 0.0000-0.1545-0.0283 0.1550-0.0258-0.0394 0.0576-0.0002-0.1848 0.0268-0.0039-2.9635-0.9414 p-value 0.0000 0.0000 0.0000 0.9920 0.0060 0.3740 0.0000 0.0000 0.1580 0.0260 0.9660 0.0000 0.0000 0.0920 0.0000 0.0000-318.557 Macinery and equipment 0.8742-0.0492 0.0827 0.0054 0.2018-0.2467 0.0387-0.0335 0.0550 0.1492 0.0077-0.1812 0.0230 0.0003 p-value 0.0000 0.0080 0.0000 0.2430 0.0000 0.0000 0.1250 0.0000 0.0360 0.0000 0.0540 0.0000 0.0000 0.8830 0.962 Electrical and optical equipment 0.9492-0.1095 0.0674-0.0013 0.3057-0.3006-0.0433-0.0246 0.0161 0.2595-0.0020-0.2241 0.0280-0.0029-1.3952-10.6632 p-value 0.0000 0.0000 0.0020 0.8060 0.0000 0.0000 0.1010 0.0000 0.5470 0.0000 0.6550 0.0000 0.0000 0.2780 0.0000 0.9140-411.175 Transport equipment 0.8715-0.0388 0.0587 0.0143 0.3866-0.2692-0.1374-0.0265-0.0309 0.2291-0.0100-0.0444 0.0244 0.0054-5.5180-0.6306 p-value 0.0000 0.0020 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.1360 0.0000 0.0000 0.0530 0.0000 0.0010 0.0000 0.0000-272.586 Oter industrial products 0.7074 0.0076 0.2406-0.0056 0.1167-0.1757 0.0479-0.0227 0.0656 0.0074-0.0078-0.0219 0.0302-0.0015-2.6278-0.9075 p-value 0.0000 0.6870 0.0000 0.2460 0.0000 0.0000 0.0940 0.0000 0.0310 0.7510 0.0520 0.5270 0.0000 0.5170 0.0000 0.0000-360.393 Source: prepared by te autors, on te basis of information from te World Input-Output Database (wiod, 2012). Note: * estimation by ordinary least squares (ols) reports adjusted R 2 instead of te lo of maximum likeliood. A comparative analysis of productivy in brazilian and Mexican manufacturin industries Armênio de Souza Ranel and Fernando Garcia de Freas