W PD 1 8This paper is prepard for staff use and is not for publication. The view are those of tha author and not necessarily those of the Bank. INTERX&TION&L BANM FOR RECONSTRJCTION AND DEVELPNENT Economics D4pcrtnt Working PapGr No. 13 PRDUCTIVIT OEI!TH AND DEYE PN IN LATIN Ak=Ck Nmy 31, 1968 This poper was prepared by Professor Henry J. Bruton of Williams olDege during July-Augut 1967 thn hq wrked as a visiting scholar with the Industrialization Division. It builds upon previous rasxarch done by Profesaor lirutton, ooza of ethich is also reflectod in his Dacember 1967 article in the A=arionn Economic Favio on "Productivity Gro"th in Latin Amarica." The papor rocusen on the dorivation of the "production function," i.oe., output an a fntion of, on the one hand, increased inputs of capi- tal and labor and,, on the other hand, pr.ductivity grocth. This loads to a comparison of productivity trnds in industrialized countries and in four Latin Amarican eountrieo. Latin Anrican pr.luctivity groith oneral- ly is found to be quite low, and tho author tries to deduce (partly through statistical teocnique) comsa og the major asons for this otate of affairs. In a final chapter, so= policy implications arm sugested. As Professor Bruton himalf ophasizes, manny of his findings are heuristic, i.e., their lies mainly in their inducemant to further research. In part, the difficultios am statistical, e.g., the unreliabi- lity and ambiguity of existing output data. In part, they am analytical and reflect our limited undorstanding of the reop.ctive roleB in dyamic growth on (a) resource allocation in agriculture and industry, (b) volums, timing and efticiocJy of iarwtructumr investmxnts (including ependitures in edacation and reaaatch), and finally3 (c) the political and social frame- work, Professor BEaton viahes to acknowledge the many useful coents on earlier versions of the paper by Hertil Walstedt, Herman G. van der Taks, Bela B&lassa, Rarique Lerdau and other inbere of the Bankt's staff. Industrialization Division Prepared bys Renry J. Bruton TABLE OF ONTENTS Page No. Introduction ...............00 . ........................ 1 I. CDNC AND ME&SUREME.T PROBLEMS .................... 2 A. The Moaning of Productivity Growth ................. 2 B. Choice of Productivitr Mesaure . ... ......... 4 IT. GROWTH TREN16 IN PFDDUCTIVITY ....0...00............0.. 7 A. InduBtrialized Countries ..... . .* ...... ..... ..O... 7 B. Latin American ODuntries ............0000 00000000 13 III. INTERPREING THE LATIN AMERICAN EXPERIENC ............. 15 A. Productivity Growth and Import Substitution .. o 15 B. Productivity Growth and Emports ..........0o..0000 18 C. Productivity Growth and Structural Change .0.000000 .19 1. The Nature of Strmctural Chan4e ................ 29 2. Sons Pindlngs from Latin America ............... 21 3. An AtteVted Interpretation .................. 28 D. 9umry .................o... .......00. 32 IV. SOME POLIC' IIKLICATION3 ................ 0000. 0 32 1. Size of Inveatenta in InfraBtructure 0..0.00.0. 33 2. Erficiency of Inveatmente in Infrastructure .0.0 33 3. Difficultise in Policy Changes 0*o.o. .6.. 0.. 33 40. Resource Allocation and Productivity .0 000000.0 34 TABLES 1. Annual Average Growth Rate of Total Uutput and Producti- vity for Certain ODuntrieB .... ....00..0.0.*.... *, ,0 8 2. Annual Average Growth Rates of Total Output and Inputs in Certain Latin American Countries .................... 3. Relative Variation in Selected Serieo *..............0.. 23 4. Relationship Between Earnings per Worker (W) and Productivity (P) for Selected Countriet .... 24 5. Relationship Between Unit Labor Cost (ULC) and Productivity (P) for Seleeted Countries .. .0.0. 26 Page No. 6. RelatioriH±p Bwetn Pmdtwtivity (M?)" 0 Qautfut. (Q) for Se1ects& t=trie. _. ... . . . 27 7. FPlationship awma &Tamt- (E), atS-i# (,Q) in Selecte Co *.t.... .... 29 APPENDIX: STRU=RA.L MXNG& AND PR=== 1 TIN; AMIR GA STATISTICAL. FTNIM MM, DErA ' WIT$ PlD1UCTIVITY GROWTH AND DEVELOPMENT IN LATIN AISRI CA Introduction Over the past decade a large number of studios haw been presented that shot that long-run increases in output per tsorker cannot be satisfactorily explained simply in terms of rising capital/labor ratios as these ratios are conventionafl,y measurad. The accumulated evidence covring numerous countries and a variety of tins spans has convinced many economists that additions to the existing stock of machines and equipmeat identical to the exdsting ones can explain only a very small part of maa- sured incroasec in output per workor. Tho evidence applies to individual soctors of an oconoU and to total output as wol. As a. consequanco of those findings tho rolo that capital fonmation played in growth aad develop- sc.nt theorios in tho late 1940O' and oarly l950's has been considerably modifiod, and attontion is now being directed totard other possible sources of increased output. Increased output in excess of that eqlained by increased inputs is knowm by numorouB aamos - the Residual, a Noasure of Our Ignorance, etc. ° but psrtuips tho most co aron (and the most euphonioua' iz sizplyl increased productivity of tho eonwntionally identifiod, i.e., labor or capital in- puts. The present paenr is one of a series of studies liiich seeks to Qxamine various aspects of productivity growth in several Latin Ase.-ican countrios9 Althoug primary cor.::-n in this raport is eith a single, and rather narrow, aspact of the general question of the role of productivity grouth in Latin Amarican developtant, it is usoful to see how this particu- lar issue fits in with the other issues treated in detail in otbh.*r papers. In Chapter I, thorefore, a brief reviou of a variety of issues is presented. It also soard useful to include in this chapter a ahort discussion of some of the conceptual problzes aswociated with offorta to measure productivity grth-. Chapter II analyses grouth trends in roductivity in ind-9trialized countries and in Latin Amrica. In ChapterIII consideration is given to the role of productivity grouth in bringing about ehat is called "structural change.,, Finally, in Chapter IV, an attempt is made to explore some of the policy impJicationa of the findings and arguments revieved in Chaters I through III. Ciater IV ziy be read first to got a general notion of the paper. I. COONCEPTAL AND I'SkURMNT PRDBLTE 1 This chapter has ttbo objectives. Section A raviGus the various bays that productivity grouth czn be defined and msasured and the kind of evidence that leads to the conclusions that capital's -role in the growth process is other than az° -thooght to be the case in the early days of growth md development ecoanomic. Sactlon B-revies a variety of aspects of pro- ductivity growth that have beae treated elaethere and that are ralated to the central question examd in C&pter II. A. The M9sning of ProductivitX G I/ 2 ViaTe rats of grovth of productivity try be defined as the differeLnce bateean the rate of growth of rcel output and the rate of gomvth of real input. Write P for raal output, ? for input, and A for productivity, then A N ' N tDre A/Al & P/ PS tc. ara finit2 cPro imatios to the rulativ xates o± grouth of tk Undicatcd varIcb1m. b7er interest is in an %gregage output, emd Lora thbn =a inpat is isolated, e.g., GN or mmufacturing output, than the dafinitione zy bea mItUm as &A 2.~0j i - n ytk p, i.N t&g t and v am relative s of occh -pE:ct and input in ragate output and eggrggate input roqpeclve2y,R2/ 30 > Thes definition vp a voriety oi-iswias relevant to an imssstigation of productivity gxwuth. (a) Tho central notion ia clear enoigh. Imputs of givn chcaracteris- tica and quantity exist in thG two period and-produce a specified output. Over an identified time span if the quantities of inputs all increas9 at tha saim rateg and all characteijritice of tbe imputs Greain exactly as they vare in the base period, then ou,mt and inputs grow at the sme rate and there is no change in the productivity of tJie inputs. If output increases =Dre rapidly thaa inputs, then p ductivity asa by definition increased. Evidently therefore, onie mut be hle -to eaure the quantity of inputs indepandantly of their value (i. independently of their productivity). 1/ TBe reader conversant with -tt xrecent litsmrture on productivity groath may wish to omit-this section and tha following one or to skim them quickly. 2/ Furthar elaboratiom of this procedure iu in D. W. Jorgenson and Z. Griliches, "The Elanation of Productivrity Change,, i_view of Economic Studies, July 1967. - 3 - This latter problem has attractod . -3at deal of attention with respect to the noa3urement of capital as an inputt, but it is important to recog- nize that tha problem e2iats for any other input that one may isolate. Unfortunately tbhra ia no complotely satisfactory way of so measuring the inputs. (b) Thera is generally an aggregation problem with respect to both output and inputs. The usual practice of deflating an aggragate output series by a comwon deflator introduces an error. So too does the common practice of summing input services in constant prices. These sources of error camnot be eliminated without rather more complete data on the com- pononts of the series than are usually available - especiaUlly for the Latin American countries. (c) A ccrollary difficulty is that of separating actual inputs of services fro.n available stock:j of the Inputs. Again the capital input is the most obvious problem, as capital stock is evidently fraquently avail- able and not used. The sama problem however exists vith respect to labor and any other input that one may identify. What appears as rising produc- tiFity may in f.act be simply increased utilization, and one must therefore make assuiltions as to the changing degree of utilization of the factor in- puts. (d) A further point of direct relevance to the following discussion has to do with factor prices. This problem is the other side of the quantit7 of input problem mantioned as items (a) and (c) above. The idea is to measure the prices of the input services rendered and to be able to assunm that those pricos represent thc -arginal productivities of the inputs. Both the measurement and the assumption of equali ty between price snd productivity are frequently defoctive. About the best that can be accomplished with present data is to aBsumo errors are about the same over the time span being considered. If this were not the case then the conputed AA/A would reflect changes in the quality of the measures as well as in productivity itself. Again, given the availability of data, all that can be done in the present context is to alert ourselves to the question and, to the extent possible, qualify interpretations where available evidence indicates price figures do not measure exactly the same thing over the time span covered. (e) A final general point has to do with the quality of the da.ta. Data in this field are always: adequate and incoriplete, but this fact is not a sufficient reason for ig ring them or for proceeding to use an unsatis- factory model simply because good data are available for it and aryj not for a ra'e powerful model. One of the important attributes of good aralys.s is the effactivene5s with which available data are used in the conteo:t of the mosrc suitable model. An irnportant task then is to use the data wkich are availabLe in ar effective m-iel to see what t.ey show. One must then be equipped to appraise the results in terms of whether they are significantlj affected by the shortcomings of the data. But this is a much more useful approach than is t,e use of less powerful models (or no formal mzAdels) with bette. data, and tryin^g to decide the extent to 'ihich the resiJts aie defective due to the model - or to what the data mean when they are preaented with no model to guide the analyst in an inte?pretation. - 4 - The prceding are all important points, and an inmortant argirnent can be built around them to shob that much of the resear'h that indi cates a modest role to inputs in accounting for groi is dubiouB (see especially the Griliches-Torgenson paper cited in footnote 2). At the present tile, hoiever, the evidence is still clearly in favor of the position outlined in the opening paragraph of this paper, and it is deemec uselful therefore to measure productivity growth in the manner described below and to seek an explanation of its behavior as tell as to examins its consequences. B. Choice of Productivity Maasure There are tuo Nidely used measures of =%xiuctivityz the average pr3duct (gross or value added) of labor o.r the average product of a capital and ldbor input mix. Which maaEure is momr suitable depands, in large part, on tho problem under considera ion, end ono shuuld not think in terms of one mscnur as generally superior to the other. (a) Coneider first tho -z VnuSant of the .'pvth of productivity of th3 capital-labor combined input. The naval approach is to iurite a produc- tion function something like 1) Pt - F(7t Lt St) Hora 3t oad L, raproseat the qumntity of ep-4.tal and labor e-rvices available during timn pariod t9 Pt is gxoae drzSatic (or national) product during the o= period, and St in a hiPft fator.j $bh interpietation of 1) is quite Ginple. It oee6rta that in yarod t the mnzam= output tho acononr can pro- &ao io datkazui=d by tho qrmtity of ecpitaL und labor oer'ice available =tS '9n aatvzo of t-ba-Ar xaJnddln-k as given by F. Th3 ohit ?.&ctor S iagicaWa that vovr jacowsm the eutixi'ftn&Aon a in sucn a E r that direct izmutnB I nd L bae (coneivably aezingly) prodLcti±v, i.o., P grow =o rapid:y ts tho tighted average of ccsital and labor. It z!3 criea tvidho shift factor to bo of equal or greater irport&ce thai re&d caital and labor that lead to thea ohasio on incmrea in produtivit no m egsential eP2mnt in groarth thsory. So tho search bi ra ar fo ezplnation of S. 6. One fraquently used aZpro#Cb io tbG folloving. Sq)poaF there is a hift factor acting in such a mWn that it has no effect on the iarginal rats of subatitution betwe1 capitu,) d labor, but si;I.y incases (or boraases) the output obtainod frog th given iruita of capital nnd labor. In this event the function ixoe mW be wittX as Pt - A f(Kt Lt) cad At masurea the c latL'ue wit that oeet mer successiva periods of ti0. To find a waV to isolate A requirs further specification of the form 3/ amr a also factor out 2abor and put t gwent in term of outp"it per corkar, capital per varkar, and shift couponent defined in terms oi output psrvorker. of the function. The oat frquently used function is, of course, the Cobb-Douglas production function written as Pt ° A KatLt a + b l 1 whero a and b are elasticities of output with respect to capital and labor respectively. That a + b a 1 means, of course, tnat constant returns to capital and labor arm as--=ad to exist. Now take logarthma,, differen- tiate totally with respect to time, and assume that changes in A do not affect a and b, then AP/p AA/A + a L K/K + b AK/K and evidently AA/A - AP/p - (a A K/K + b A L/L) This last oquntion states that proportionate increases :ln the productivity of capital and labor are equal to the difference between the actual change in output and that which would have occurred had there been no change in A. F&qplainL-: productivity change then means explaining &A/A of a particular time period.kv 7. Although the preceding m=asure of productivity grouth was derived fromT a particular function, the form of the function does not seem to matter a great deal insofar as estimating AA/A is concerned. Another common function is Pt Ct (hi° Lt + ro Kt) t L + ro Kt where w0 and r ar payments to labor and capital respectively in the base period, and C then is the measure of productivity growth.5/ The measures C+ and. At a'ill not yield very different answers for most data.6/ The form then of the function does not seem to be of great importance, Sut the other assurptions - zDecially constant returns to scale - are and relaxing them will give different estimates of productivity growth. 4/ The argument produced here was originated by Robert Solow,, :Technical Change and tho Aggregative Production Function," Review of Econorics and Statistics, August 1957. It has been discumsed ata leneth mny others. A good rzviv-ew is Richard R. Neleonle "The Aggregate Production Function, " The American Economic Feview, SepteLiber 1964- 5/ This formula has been used by John W. Kendrick in M; 3 study Productivitz Trends in the Uiited States, Princeton University press, 1961. It has also beer. used t,y Imany oUEers as well. 6/ The same secms to hold true for another newly popular production function, the constant elast:c-ity of substitution function. (b) Prior to the last decade or so, prvductitVtty growth a-most ala.ys -mant labor productivity on.y., i.e., ch.anges in tLe ratio output/labor. Interest in this rat4o 0 'L? to some extent a matter of its use as an indicator of econonic we2lare. Also it was frequent4 used as an lndicator of teic ,,total" factor productivity discussed above. Fbcently, however, its Us5 has rested on consideraTions of relevance to the pmsent discussion. 8. In the modern sectors of many activities what might be called a Ienntief production fiLnction seems generally applicable.7/ In such a function the capital-output ratio appears to be more or less similar from country to country (aggregate or sectoral ratios), sle labor output ratioos var-v idely. A more precise statement is this: the variance of capita. output ratios axnog countries is so vexry rmch smaller than the varib..nce of labor output ratio,a that it seem not i=aralistic to s.-ms that the capital out- -put istio, rolative to the lbor output ratio, is the sama for a given modern sector in all countries.8/ The. the captal output ratios (espacially full capcity, fixed capital) ara si.lar is jal-ained in tenrs of fixed coeffi- cients of production on the capital sido and ths fact that so much of modern capital equipment outside th Oo st countries criginates in the United States and Moatern Euope. With theae asuumptions the rate of growth of labor productivity becomes oapaciJlly ralevant. In particular in making -ixarnational coaparisons of rates of growth cf productivity (or rates of xnecalne in costs) attention can be 21rAted to labor productivity.9/ As noted balow this argumnt 4S u2oful in the context of understanding theirnlica- tions of an iport substitution atratr of dnalopmant. 9L A further consaibrtioan in the use of labor pmductivity has to do trth some evidence -hat tin p=oh=ti-ity of non-labor inputs moves in the samc diLection and at about the ue -ate as that e' labor.lo/ Labor praductivity groSth rW then sarve as a prxy.for the more inclusive measure. N3?zei:portantly, this evidence also asugests =wathing about the sources of epodiutivity groaths natmly that it com; to occur generally in an activity sand is not concentratGd on one imput. 'ia latter point incidentally is cmsistent vith that noted earlier, -that a rising capital-labor ratio ex- plains a very sms1 proportion of obearFed incraases in labor productivrity. 7/ Zee espocially' W. Leontief, "An irternational Comparison of Factor Costs and Factor Use (Review Aticle)," Mrrizan -Economic !4viev, June 1964 8/ .Some frieads in the World. nk -hime trang3y challe ad z =orking k1pothesis that the capital-output ratio tend to be the sam for a :givan dora sector in all uontzivs. 9/ -See Bela Balassa, "An mirical 1knonatration of Cassical Coarative -ost Theory," Review of Economics-ad Statiwtirs, August 1953s and -borah Paige and Gottfried No3acH, A OoMarisoon of National Output .zmd ftnductivity, OEEC, Paris 1959. 110/ .-.a in W. E. G. Salter, :Prm e nd Tmbniza Clzt;ridge 1960 -and the Kendrick volu cted in footr-ote 5 suggest this to be often true. Aors on this point in Chap-ter IIM - 7 - 10. n:ere ara a variety of other issues associated with the pro- duc>'*ty concept and ito cn&juzvre=t. The preceding brief survey seemz adequate a&i a means of F-king clear the basic definition and the m.ain approaches to mea'suremant.1l/ II. GFOWTH TRENDS IN PRDDUCTIVITY A. Industrialized Countries 11. The earlier theoriaa of grouth placed capital fonnation at the hear". of the vmvLh process. Note has already been t&ken of the evidence that measurad productivity growth is se great that it implies that something imors (or other) must be at tho heart of this process. The purpose of this section is first to look briefly at soms of this evidence, and secondly to derive a more satisfactory view of capital's rol3. (a) Table 1 premenLs estimates of &A/ for GNP for a number of countries for which data ara readily availabte. The last column in this table shows that productivity growth as defined by &AIA accounted for .'rom 38 to 79 percent of tha growth of output for these countries during the 1.950's. In most cases the percentage exceeded 50 percent. Evidently theni a theory of growth that leaves A A/A unexplained, leaves about one-half of greuth unexplained. 12. A similar pictura emerges if one looks at individual industries or sectors o! the economy. . .:_ Kendrick data one may compute that proportion of grosth of output dus to increased productivity (defined by C above) for tbs two-digit ISIC industry groups in the United States. These proportions range from a lcw of 17 porcent to a high of 74 percent for the period 1929 to 1953. The mean value is about 40 percent and sixteen of tienty ratios fall between 30 and 60 parcant. 13. Evidence present3d by Reddavgy and Smith for thirteen manufactur- ing activities in Oreat Britain for the period 1948-54 indicate a picture similar to that described by the Kerndrick data.12/ Their data indicate that from 12 to 32 p3rcent of the increment in output between 19L8 and 1961 was due to thle increased productivity of capital and labor. The aver.&ge for all ac. ' vities was 47 percent. Again then it ie -.vident that in not explain- ing producTim%ty growth, a substantial part of actual growth is left unex- plai ned. 11/ There are many papers diecussing the various difficulties and issues involved in defining and measuring productivity growth. The Kendrick volume (footnote 5) has some useful discussion. See also ivsey Domar, "On the 24easurement of Technological Change,r Economi- Jou-rnal, Decerber 1961. .2/ W. B. Reddaway and A. D. Snith, ,Progress in british Xaznufacturing Industries in tae Period 1943-5$1," Economic Journal, March 1960. TablQ 1 Anum2ai Avore Oroutb Rate of Total Output and Pr-odc iiyfrCrain Countries CDuntry and Poriod A 4A P/P Belgiumn 1949'54 3.6 2.5 .69 1954-59 2e3 1.6 .70 trtherlnnds 1949-54 4.9 2.7 .55 1954,-59 lhol 1.6 o39 19149-59 3.7 2.3 .62 I'zaden I 9 DYt. 'V 3.4s 2.5 .73 Unitod Htingd IA9-9 2.5 1.2 .1a -.8 3.8 .79 19!k.59 lgl 2.8 .68 9t 95 6.4 O69 IM-51b9! 507 .o 72 1950)-51$ 0.3 5.6 67 195Io59 6.6 305 .53 Isra-al 1952.58 9.e 3.9 040 Japan 1950-58 7.9 3.0 .38 Vaitod Statoo 1947-5;4 4e 2.9 .66 195-60 3.5< 2.1 .6c stmves a Economic Comdanion for Euopes, Sam Factor' the Economic _zovth in D=" Ring the 1 s ted NAtione, Geneva, 1964o 0. Ankrua%l "Factore of Eonomic - A eview of Bacent ?eaearch", Prodactyi-e Ret eview, P6b. 1965. R i hrd L 3. blsons paper cited in footnote l. M17 J. Buton, "ProductiYity Gzwth in Latin Amrica," American Boanouic ~vi Duoelr 1967, 11a. ;i is useful to note also the wide variation among the sectors in the rate of growth of productivity. This wide variation holds also for the United States data, anri, as noted below, holds fo- some Latin American countries as well. This fact has important implicatio.s for au.y explanation of productivity growth, and for a variety of factors asoociAted with some distribution and allocation questions of economic developren1. Some of these issues are explored below. 15. Data for the less leveloped countries are not so plentiful, but one Lmay offer a bit of evidence even here. From data on rates of growth of GN? (or GDP) for the decade of the 1950ts and for the average investment GNP ratios for the same period for 39 countries with per capita incomes below US$500, a least squares regression equation relating the two variables may bo computed. For 39 such countries the equation is a P/p - 2.02 + .'8 I/p r2 + .23 (0-5) The equat.ion indicates that if the investment rate were increased from say .10 to .20 the growth of output would increase by about 50 percent. In the simplest version of the capital output ratio model, on the other hand: a doubling cf the investmer., rate will double the rate of growth of output. 16. Ir. the sane rough and ready fashion one may interpret the intercept figure - 2.C2 - as the rate of growth of output not related to capital forma- tion. The average annual rate *,rowth of output for these countries was h.95 percent, so about 40 percent o; the actual average growth of output would have occurred had the investment rate been zero. In general this latter figure is doubtless too low as variations in the investment rate explain less than one-quarter of the variance in ANP/p.13/ Despite the obvior-: short- comings of the argument, it seems safe to say that this bit of evidence is ^onsistent with that for the richer countries, namely that capi+al formation, as conventionajjy measured, is not alone in its growth producin, powers.li/ 17. Much more evidence could be offered, but the preceding paragraphs should give a flavor cof' the kind of evidence available to support the posi- tion that an explarnation of productivity growth is an essential ingredient of a thTory of economic development. 'b) Evidently these results have major policy implications for deveiGping countries and for aid-giving countries and lending agencies. It is urefui. then to try to say somethiig about these i.plications at a general Invel berore proceeding to 3ome more specific issues related to the producti- vit'v question in Latin America. 13/ whe wald surely be higher if actual data for the rate of growth on ca. tai were available. This is sug-ested by so;ie data on actual rates of growz-i of caplt_a. znd output. P.' ; sirxijar -ton for il low :.n,z±.:> countries was computed for 'he period 19--6oi. The resulting4 equation is NP/p = 3.32 + .10 lb (.08) ana . s 'ri tunZ . ,lr this short pe.iod however there are al-. wn-, 'L . eaa-ia. ory power of the denent variable s ow; . t'.a for_ :-..c _ .:3 weur "ea froml the `'o:-ld Tabs of L. ._L,. - 10 - is. Obviously we need to know something about the sources o2 o-.,- ductivit, gro'rth. As long as a production function with two inputs. capital and lahn-- and constant retFrns is used then these sources of productivity growth are limited to inprovements in the quality of these inputs and in the efficiency of the system in the allocation and utiliza- tion of existing resources.. Consequently imorovemants in the quality of physical capital (new technical knowledge "embodied, in plant and equipneat) and imDrovements in the quality of the labor force (education, health, exe.) have received great emphasis as source8 of growth in numerous studies 'lo date. Similarly, considerable attention has been given to improving the allocation of resources and to eliminating underutilJLation of existing c api tJ. 19. It is imiediately clear, boeaver, that neu technology, increased education, or, indeed the desIgn and execution of policies to effect a better allocation of existing rsources ara not frae goods. They can be acquired or created only at a cost. This cost is in the form ofr the use of currently available resources for purposes other than those bwhich rield immediately available resources for purposes other than those wiich yield immediately arailable consumer goods. In this sense their acquisition or creation requires saving in the sams sense that the building of a machine to make a machine requires saving. Thus it seems reasonable to speak of "investment' in elucation, in research, in health, and in improved alloca- t.on. Once it is accepted that these sources of productivit growti are not frae and that they do y^e.Ld a flow of serVices - i.e., their existence does result in output being higher than it tould be in their absence - then the sOurces of these floms arm a form of capItal, for the essence of caDital is the fact that it yields a atream of income over time, and that indeed income is the product of capital. Furth3rmore, as the creation of any form of caDital involvs,, a cost, it is socially morthuhile only if the capital value of tha income stream it yields is greater than the cost of the capital. From the recognition of the invastmsnt charcteristics of this type of expenditures it was a quick stop to seekLn% to reasure the rate of return on such investment, and to seek to idantifyg that part of the groath of out- put (total or per worker) that could bo attributed to these various factors. 20. In a general sense than, grvath and development theory moved fromn a narrowly defined capistal approach (in the late 40lo and early 50's) to * som-3what anti - or non-capital approach (in the late 50's and early 60's) to a vide all-i.:lusive caDital approach 'as of now). One may then think of the centrs. task of development as consisting of finding investible resources - i.e., resources not required for the production of consumer goods - and to use them to creats that form of casital, very broadly defined, which will make the maximum contribution to increasing the capacity of the econoMy, We may speak of this as the generalizr- capital format .on approach to development.15/ 15/ See the short "Comnent. ot Harry G. Jc.knon in 7he Residual Facto. and Ecoromic Growth OECD, Paris 1964. g'ni p'neraTized capital formation aproac underl-es the app-roach. followed in Henry- J. Bruton, Principles of DeveloDment Eeoaonocs. Frintice-!izU. Zg2eivwoo CLLffs, *'Jew Jersey, 1965. - II - 21. Little real progress has been made an empirical work "' their contex-t of this approach. 1/ Of the manI difficulties involvea -n an empirical _nvest2.7ation there &re three that are e-?ecially pertinent to the present discussion. (i) Some major source of productivity growth may have been overlookad. % long list of suggestions is possible, but one source that seems very important for developing countries has to do with learning. If a new activity is created in a country, then one might expect that pro- ductivity of the resources engaged in this activity would be relatively low at the outset but rise as management and labor became better acquainted with what they were doing. In this case the increased productivity requires no outlays on Aducation, research, health, etc., but does imply subnormal profits or extra high prices over a period of time. Empirical studies of tAis source of productivity growth are scarce 17/, but some generalizations seem possible. it cannot b7ea continiuing source of increased producti- vity. Ir. most developing countries, for example, productivity in traditional agriculture is not going to rise because of a learning effect. Even in new manufac- turing activity the few available data suggest that laborers and managors cga learn specific tasks quite readily. This doee not mean that learning is not important, blt rather that - ..,)act is more or less dependent on the continued emiergence of new activities. Thus the extent to which learnL_ag is importaDt depends heavilZr on the capa- city of the econorm; to generate new opportunities and new aituations .18/ (ii) The role of technological change, embodied or otherwise, appears the most intractable issue connected with produc- tAvity growth. ror those countries currently using a primitive technology, borrowing from abroad appears a rea:soxiauIlt: approach. Thei-u are, however, wel-ll-ovni difficulties associated with the suitability of a tech- iioloa c-eated for one economic envirormrat for use in a different one. Similarly little success has been gained 16/ Thei most a'nbitious enqpirical effort has been that. of sdward Denison, Tile Source:i of i.minnr.z urowc7 in the Unitted States, Com;imttee for Eco;nnmic levelopmrC.e, New York, 196-2. This attempt is inte:estiniy aid imaginatively done, 'out the results are more illustrative than real. 1;'/ hNie best evidei.te. has zo do with exper'Lences in ccrtaia nlew acwVvities in the United States du-nag the World War TI. See, for example, Leonard 9a?pi.ng;, 'Lear- n and World War II Productioa" Functions,' 1heview of Ecor.v)mrct .snd 3tatis.ics, Februarl 1960, Ž6/ See the foilowihg fae Sor this footnote. in identifying ur measuring returns to resources engaged in technical reaeaenh. In the context of development. perhaps the single most imortant issue is that of ,borrowing technology from aboad' or 'creating one's oum technology'. The resolution of this issue Ls essentially emirical and, ae just noted, littlo h i'ay has been made on Gzpiricel investigations of the quest±on as of now. The usual approeh at the preseent t1s is to san suppose that output groz at x psrcent pr year becau,e of purely tooeaolaTLcal fctors, thon see hov such an aseuztiou fits in vith soma of ttw data that we do have.19/ Most observers seem to think that technology is of gr;tat raiovaence in productivity growth, but nobody cm give wory fira ev-idanoo on how and h:ru auch (iii) a third iesw of cpiznJa lom tco In th3 present context and for polickv matt= tnll hoo to do with tha inter- d6pandenrce of th vmriv" (eid) sources of increased productivity. Argoznt ~d oupirical Pox ofte isly ,&at the effect on productivity of (agy) edcation and te.hnological chmn,p ar- aiddtim0 Evidently thio is not the em. A tachnologir2 innoation a rsquira new labor sklls b2fsro it ia effective. Ora increased oduoatic-n m-y rtlt in nGr tochnical kb:lsdg. And so on. Thvs to identif'y oducnt±on on a source of producti- vtjy goth a fx%m t praodnd by techoological eb gs 1i toe thie nawre of thoem sources Tho second consnE~e of ty ia n of K imdliag welev=, r s Am at contrmn to that Jiat-noted. :3ipom It tere grsd t; a division of the z cs of 2 t productivity growth into increased odueatia kV t.i ID e~o Damiscu sd othetes e=cpt1ly isnpossible b.crause of the coilemnt=i-W p It is not possible to gy, for ezale, tat c-thz'd of tho growth of MMIP in country X o t it 50 yrc ha ito Bource in od=ation. Do"es t.in the Cneralised cSipittel formation approab to d l , th au at tha end of the pzending sction is invalid? Suzely it does not meo t It la o= tIdi'to dissct tha pat,, d another thig to o about to future. Evun 18 on pe 3s11: U=Z% Arrow, "The Econeac I A1icuMon1 of Iemrni%ng by Dibng," Rmis of EConomc Studi. Va. ? l96L is the classic treatment of ErD d g pro8e8s a w of P T uS titVity grovth. Arrow for tk,2 z3t part ma lsa.rrd a f on of ccumulated gross invest- xt oa the groua that the ltte.r ia effective indicator of the sWars of n A-hi md opportSmties for learning. 2i/ fv' v -22wSn pZr ciEd in fo3tnot-; 4. 13 - with compl % esatrity it is still possible - indeed necessax7 - tm ask, given the existing situation, wil. expenditu,m c. innvestible resources on education - or technical rasearch, or hoalth clinicss, or increased consumption - yield a teturn greater than expenditure elseuhror. This is a msaningful question to ask and one that is and must be anasered by the nolicy maker. ThuB our ultimate interest is in the allocation of investible nrsources among the various capacitr increas- ing uses which are open to the society. We learn about the consequences of in7estments by studying what has happened in the past - even though such studies are necessarily defective. This then is the logic of the emphasis on sources of productivity grouth and on the generalized capital formation development model. It is important to emphasize that thia model puts heavy weights on saving (capital formation) and on productivity growth and than broadens the allocatiZI question in a correspond- ing manner. This approach has evident policy implication for both the developing countries and for those lending and aid agencies which hava .a concern in development. B. Latin American Countries 20/ 22. Efforts cere made to estimate &A/A for Argentina, Bra4il, Chile, Colombia, and Mexico for various time spans since 1940, The rasults are summarized in Table 2. The mean -alue for L6A1A for the twenty observations is 1.4 and the mean of :nA/A as a proportion of A P/ is .26. Both of these figures are significantly smaller than the corresponcLng est4 mates for the countries iii Table 1 (3.0 and .60). Correspondingly the ruol of capital formtion (narrowly defined) was larger for the Latin AABrican countries. Two furthor reaults ara of special importance. Productivity growth in the Latin American countries (LAC) appeared much higber during the first two periods sho-n than in the second two, while the rate of capital formation was considerably hi,her in the second two than in the first two, In a general sense then it seems correct to argue that in the later periods pro- ductivity growth was replaced by capital formation as the major source of the growth of output. 23. A second result of even greater importance. For the advanced countries (AG) of Table 1 the assumption of full utilization of reources is fairly accuratoe (?or the United States data are computed on the basis of ,potential output" rather than actual output, thus reducing the effects of underutilization on A A/¢. See the Nelson paper cited earlier.) The estimates of AA/A for AG ion seem to represent genuine increases in the productivity of capital and labor. For LAC on the other hand, underutiliza- tion of both inputs was of great iaEportance. In this case then AA/ zrWasures the extent to which the econony is able to exploit its avaiLable 20/ See the Bruton paper cited 'ts a source of the data in Table 1. Table 2 Annual:. Ae!o Grth Ratus of Total ou!ut and Ipts i CarLi Latin American Countries A p ~ tA- ___-A Co uM and Psriod- P A .;P Argentina 1940-4;5 2.9 i.6 .55 194a651 3.4 .14 .12 1955-59 1.7 -.6 o.35 1960.44 L 2 -.6 -.50 E3rasil 194045 312; 1:.3 .41 1947-53 5i6, 1.6f .29 1955-59 5.6 1.7 .30 1960-63 5.0' 1.2 .24 19.0-45 2' 7 1.D4 .52 1946-53 3 9. I.4- .36 1955-59- 30. .1 .03 1960-64 4.0 .9 .23 1905 2A8 1-1 .39 1963. 5.25 Z. 3 . 44 1-9559 1o:10 1964E 5L . 33 1940-45 9MD 6e 7 .74 194653 5;0) 1-.2 .24 1955-59 572.0 .35 1960-64 q.22 Z.8 .45 Source: Bruton or c.d otw r - resources as well ao the kind of "tgenuine"l productivity increments discussed in the previous section. By a series of regrassions involving data for various time periods, results are obtained uhich suggest that virtually all of the variation in AA/A after 1955 is accounted for variations in outp:at, i.e., "geauine" productivity growth seemm to have fallen to virtually zero. 24. That productivity growth in LAC seemed much higher during the war years and immediately after than it has beea over the last decade or so is an important result. During the war years real. capital formation wa deci"'edly hanperad by the limitation on imports. Thus the frequently en- countered argument linking productivity growth to capital inflow is not supported. Indoed regressions of &A/ on t-h rate of growth of capital anrd on the growth of a combined cepitZ& and labor index indicates that for neither LAC nor AG arG the direct inputs a vary important carrier of produc- tivity gruth. 25. The explanation offered for these results as to productivity growth in LAC rests on three major points: (1) a groving inappropriateness of the input mix in production due in large part to internal policies with respect tr, exchange rates, interest and wage rates, and monetary and f-iscal matters; (2) a growing inappropriateness of the composition of output in the sense that productive activity was not baaed on cost or potential cost con- siderations, but rather evolved in reference to the incentives generated by protectionist policies designed largely to meet balance of payments crises; (3) a decline in competition. The 3mpi.rical results and the explanation offered for Whem suggest that the import substitution strategy currently popular in LAC has created an ei _1Jc environment inconsistent yith continuing growth of productivity. At the sa$ time this strategy has created numlrous gaps in the capital atructue that has produced obvious investm-ant opportunities These opportunities havo been exploited, keeping tho rate.of caq.ial fo ation genorally quite high - despite widespread underutilization of midsting physi- cal capital. That this sequence of events had and has now considerable relevance in LAC's devolopment difficulties seeme evident, ar- is commented on below. III. INTERPRETING THE LATIN AMERUCAN ExErCEENCE A. Productivity Grcrth and Imort Substitution 21/ 26. 'here are rb.y sides to an import substitution strategy of develop- ment. T..3 paper referr9d to here cLuis4ders only one side, that havin; to do with productivity growth. The centr:L elemnat in Import substiltution strategy is the protection (by one meaws r--: another) of certain activitiss from foreign competition and their consequent growth within the country. Evidently, the fact that some form of protection ig necessary means that the products of these activities cannot be produced as cheaply donestically as they can be abroad. .hat then caa arn import substitution strategy hope to accosmlish? One wa;, of tnkind., about a "isucces3ful" strategy is in terms of the rate of growth 21/ See Henry J. Bruton, "Prociuctiv'ity Growth and import Substitution," forthcomngu in lTe Journa.L of Development St;:.ies. - _6 - of productivity in the import substituted act;iVty. The argument can be sasily made. ,'7. Suppose that productivity is groving more rapidly in the protected activity in the developing country than it is in the country currently able to produce at tbs lower cost. At some point in the future, therefore, costs (and prices, if price follows costs) in the former country will un fact be lower than in the latter country. Thus if an activi y is protected where productivity grvw-th is relatively high then the protecd.ion can be expected to produce an industry uhich may eventually be able to survive without protection. But this mere fact is not a sufficient condi- tion to justify protection. In the period during which the econorny could obtain, via imports, the protected product more cheaply than it can be produced domestically, the econonl is obviously suffering a loss, i.e., has available a mnller quantity of this good (or this good and some others). This 0oss cui-1ated over the asveral periods in which it is incurred may than be thought of as an inveatzwnt in the saw sense as an investnent in a factory. The econo' is foregoing current output in anticipation of a l&uger output in the future. This larger output then is the retuni - the bensfit - on the investmant. One mry then calculate the present value of the total cost of this "invamtsnt", the present value of the return, and hemoe the "rate of return" on this type of investment. Onry if the rate of raturn so calculated is higher than that obtainable from alternative uses of investible resources, can "intvestnt in import substitution" be justified on this critoria. If activities were established which required permanent protection, then the econort wld be saddled with a dead weight from which it 6ould reap only losses. 28. The preceding a8rg t has a strong link with a conventional "infantry industry argum nt" for tariffs. The term "infant" seems to imply that productivity i low (oostJ are high) because an activity is small or bacaua. it is young, and consequently costa will, fall mre with increased scale or incazeaed age. This a be, but the bit of evidence noted inl 9sction A disputes it. Osta c"mot be "too hight" solely because of a scale effect, for if this ver, true then the activity could be expanded, and expoits would be (by definition) possible.22/ A pure learing effectt can be adequat.e, of course, but for reasons already noted, it seems talikely that in most instances its effect will be great enough and continue long enough to pro- duce tho necessary irc.-vase in productivity. For these reasons then it seems better to use the language of productivity growth rather than that of infant industry.23/ 29 For ezpiricul work two vsriations on the preceding argumnt are made. In the first place it i3 convenient to turn the focus of the argurent a bit and ask what dUlfferential botreen domestic and irmported costs can be 22/ Transportation costs may eocpliscte this arguernt a bit. D/ The notion of interpretA. the costs of prtection in the infant lndustry argurent as an innestnt wan suggested long ago by C. F. Bastable. See MarrVy C. KeWp, "The Mili-Bastable Infant Indusgry Dgnm," Journal of Political Econoy, February 150. - 17 - pL .ii.ted at the outset of the import substitution effort (at the begirmir.g of protct.ion) if these costs are to be equal in a given number of yearz. In this case the numb, - of years allowable would be a datum for the proiZe m, which datum would have in turn been determined by reference to such factorz as interest rates, inflation threats, balance of Dayments conditions, and the like. Suppose a period of 10 years is set, then on the basis of esti- mates of productivity growth and change in import prices one may easily calculate the allowable initial difference in costs. 30. The second variation has to do with the type of production Iwuction eq)loyed. Since most import substitution has to do with manufactur- ing act.Lvites, attentlon in the empirical work is limited to these activi- ties. For this group of activities it seemed justifiable to use what was called earlier the Leontief production function.24/ It will be recaUled that the assumptions underlying this function penis t one to use labor pro- ductivity in making international comparisons of the conrequerces of productivity growth. 31. So with all these assumptions estimates were made of the growth of labor productivity in 15 or 20 manfuacturing activities in four Latin Amorican countries (Brazil, Colombia, Chile, Venezuela) over a period of four to six years. These estimates were then compared with estimates for similar activities in the United States. FLnally, it was assumed that these calculated productivity growth rates would continue indefinitely - or at least over a period of 10 years. Then the final calculation is simply the difference in original costs that is consistent with cost equality at the end of the 10 years. 32. One co.-sequence of this approach to the formulation of an import substitution strategy is i=mmdiately evident. The policie:a implementing this strategy must be consistent with - preferabiy, must encourage - pro- ductivity growth. If, as was suggested earlier, the very low rates of productivity growth ob'erved with respect to GDP are due (in large part) to Uhe import substitution policies pursued in recent years, then the difficulties are clear. On the other hand, it is quite possible that low prodiuctivity growth at the GDP level is consistent with - perhaps even due to - high rates of growth of productivity in specific sectors of the system. At any rate we are again pushed to the question of the sources of productivity growth. 33. Although the actual computations are open to many questions, the results are not without meaning and have important policy implications. The single most important result has to do with tho variation among the sectors. For the manufacturing sector taken as a whole the difference in "initial z_& (as defined above) consistent with most equality within '/' 3oone further data supporting this approach aine offered in the paper cited in footnote 22. As indicated in footnote 8 others have challenged t he validity of Leontief Is assump tions. - 18 - 10 years i as follos Brazil 20 percent (i.e. given the calculated rate of grovth of labor productivity in Brazil and Xn the United States, a 20 percent excess of cost@ in Brazil over those in the United States at the outset of the imort substitution pregram is consistent oith cost equality at the end of 10 years)9 Coloabla 10, Thile 1, and Venazuela 20. Within each manzufazturing sector therm is also great variation, and for a number of sectors the pertittcd differance is negative, i.o ., produc- tivity grouth in the United Statos in this sector exceeds t'aat in the Latin Amarican country. In other sectors the cost differential is over 5) percent indicating a very hig rate of grocth of produc'ivity. This large variation among sectorn not only indicates soz3thing about the natuxe o prductivity growth, as noted praviou9;), but also about an approach to a successful inport substitution strateg, n&maly, a policy that protects ttivitiea irrospectiva of productivity grouth is almoSt surm to run into c 1%culties. Eually izportont is the finding that productivity growth in a Gwn sector varies fro eoatz to ont7. Part of the ta9k of an im- pwt vgabstitution str&Ato thorofoe io tla sack to isolate those activities i have the greatest potcntl2 fr ttvity gro2th. Whatever the ratioale of the protetion polics in Ut Latin arica countrieg,L5/ tge is no doubt that uowh policir-s have not considersd tha productivity pstaial o the sholterad ectivitioa. reas i,s also no doubt that mwaxy cativities have boon eroatcd by potoetim that hava no prospect of becom- i1 able to survive uithout prtection. Th ee co is that these esatrie find ttel3 ve tsh that am dros on the oconorV ra s that toI to rem&to a1 Idnk of dintmtion in the aystem that slotw = overall ggvuth in tho cmnor alliL i in the previous section. D. r-dua;MVZ. Gomwth -Ed EXOrt 3. ao omrlL to 99 of the rola of prductivity E in the imort subatitution strategy i its role in encouraging exports. 9I?Aa caw io not yet eomloto t14 emly t pointz cm ba =do hGra. In the fli;t yez= NM preLiai=7 calemIaLUmo ouggoat that pductivity growth iD Woa irportt mtory vcKiablo amwoftind for vaiation among cown- trica in oport grwth. Tte cntkm is the fairl.i bioua (but no 1rns iartent) one that a pod part of t dificulties +.bat the Latin £lrican cou0triow am harelviox g tthdx cmorts is due to their low produc- tity groth relative to tk9 ihS c t e ries. 5i 'ZThe second point ia of a =oD gonsrel xmture. Tho arguent is 2£ mt1y mads tLat countri. that e @rt (or obtain enrquited importa) ax: able to grow. The arub.w of tin a it isn simly that developing cactrias must have imports of pwyic.l oLitsl gwds in order to maintain tk2fr growth over etevdad pu2adi. Thom Is of ctuse siub3tntial evidence tbat ccwital i1ort ahoot up ap duvelopt efforts ae accelerated, 6i/TI aW sort of acoao neU umr:i-s t prlotsctionist policien a' be daWA=L 2s, Smtiqo Narcario, "P%tectioniaa and Ind=trialljAt±an in Latiu ACrieS " Bconomic BdŽtnl for La tin Aricn, Nwh 1 and . :.;_at a favorable import capacity encourages, if it does not actually constitute the major source of, growtll. ,he argument and data 5ummarzi;d in the preceding sections cast doubt on this argument. Note the followui.g: productivity growth accounts for a substantial part of total growth and capital formation little, productivity growth does not appear to be "carried" t capital formation (narrowly defined), producti- vity growth rates (capital and labor combined or labor only) vary widely among manufacturing activities within a country and among countries for the same activity, and tentative findings that suggest a significant relation- ship between export growth and productivity growth. This accumulation of evi,'ence leads to an argument contrary to that just stated: rather than a coontry growing because it can export (and hence import), the evidence suggests that because a country can grow it will export. More specifically, to the extent that a country can ureate an environmant in which productivity growth is relatively rapid it will grow and will export. Even more speci$i- cally, to the extent that a country can concentrate its investible resources on activities in which productivity is relatively high, then it will grow and export.26/ This latter alternative seems especially helpful in trying to understaiia the development process in Latin America. C. Productivity Growth and Structural Change 1. The Nature of Structural Ch2ane 36. Economic growth is rarely characterized by an expansion of all sectors at the same rate. Rather we know that the sour_es of growth pro- duce different effects on the different sectors. Growth of productivity varies markedly among sectors, zi.:_ and income elasticities of demand are different, degree of conpetition also is different, and so on. Growth then not only means increased aggregate output, but it also produces and in turn is ILduced by changes in the relative sizes of the various sectors. An important aspect of the dev-elopment process therefore is the extex.t to which the economr can respond to the demands for the reallocation of re- sources and other adaptations that are iNposed by continuing growth. Little is known about what makes an econorm flexible and responsive to new oppor- tunities, and this is not the place to explore the question in ger,eral. Some few aspects of this general problem can, however, be investigated in the context of a productivity study. In particular, attention will be given to the relationship between productivity growth and st.-uctural change. 37. "Structural change, is one of the vague but often used terms in developmrLut economics. There seems no generally accepted definition, and it will be used here to mean simply a change in the composition of output and hence in the distribution of resources.27/ The general question is ' 26/ Tlis is not to assert of course that other conditions for growth are not relevant (e.g., price inflation, etc.). The point in thne text is to seek to understand Lie role of productivity growth, not to aevelop a theory of growth. 29P neg (.92) &gvwt!nA - W727 X57p .77 .X317) . se Ti; v. _ - e) I= - th United King.iom Zf& XD 1f 1;-Dlmt 4dZ m5-3 yelor-io rspe;:- tivaj.. _ j11wa.M --in jpa renthesis under the =Ww9u -the sxt d error of the XwAr"dd= r I ll ezwo= for regession coefr.n"o - eniW $Uid i a e United Stas Am Vtr t f ad 'h ubem puted by me f=i hi. da. Latin America functions less perfect.y than in industria- lized countries (less information, less mobility, etc.). One would then expect a morm pronounced positive rela- tion,hip betueen productivity and earninrgs in Latin America. Instead w- find that, except for Argentina, increases in earnings per worker seem to bear no relationship at all to productivity growth. Theoretically, the strong corre- lation in the case of Argentina could be due either to differential grouth of skiUs in industries showing large increases in productivity or to a failure of the labor market to adjust quickly to the change, giving workers in industries showing great productivity wage increases out of proportion to the rest of the economy. Labor Productivity and Unit Labor Costs (Table 5). If earnings do not move with productivity then the cost of labor Mill fall as its productivity rises. Table 5 shows that this relationship is a fairly significant one. Moreover, the two sets of data for the United Kingdom are consistent with the notion that the longer the time span the more clearly the relationship shows. Unit Labor Costs and Unit Gross Margins. We have already noted that tb unit gross margin may be used as an (albeit imperfect) proxy for capital costs per unit of output. One tould expoct pertmp.i that higher labor productivity and lower labor costs would result in part from greater capital intensity and hsnco :->7ld be. reflected in some increase in the unit gross margin. Instead Salter found that, in the long period, in the United. States and the United Kingdom, reductions in -nit labor costs were positively correlated with reductions in unit gross margins, suggesting that im- proved management, technical progress, and-a larger volime of output would lower the costs of both labor. and capital. For the threo Latin American countries there is no such positive relationship. Neither is there, however, a signi- ficant tendency for lowered labor costs to be associated with higher costs for capital. In this respect, the Lati. American experience is similar to the short-period (1954- 1963) experience in the United Kingdom. Productvvity and Output (Table 6). In all countries exceot Brazil the:e was a sigr.Aficant relationship for the sectors studied between changes in output and changes in productivity. Analytical and stat±stical evidence advanced in the AppendLx suggest that, particularly in the United Kingdom and Argentina, this relationship was due to fuller utilization of a labs- force which, at tixes, was very Iwdch underemployed. In Brazil and Colombia one would expect similar forces to be at work since some time would nonmally elapse be;'o-e newly created industrieas achieve normal capacit; utilization. The observed po3itive rzlia;iors!iip between output and productivi'Y could also refieczt economies of scale. Yet, if economies of scale were imporMant, one would expect a clear tendency for thc - 26 - Tablo 5 Plationahip Batn Unit Labor Cost (UL0) and Productivit mP) :or Selected Countries oDuntry _ESu8tion ~ Unitsd Kingdom (1) ULC - 305.3 - .82P .82 ((.07) 'United Kingdom (2) ULC - 236.8 - .84P .63 (.12) CoolRbia ULC -1401.3 -.70P .5 (.1o4) Brasil ULC - 910.7 - 2058P .51 (.62) ArIgLodlu ULC 13o6.4-o o92P 0,u (.61) Sourot Sa Thbloo 3 and 14. Tho V2 for tho Unitod Stat4a is .80 uRing thn Stor data. - 27 - Table 6 iRelat,ionshiR Between Productivity (P) and O ut-(Q for Selected Countries -2 Cou%try r United Kingdom (1) P U113.3 + .23Q .65 (.03) Uiiited Kingdom (2) P a 72.3 + *lQ .4l6 (.08) Colombia P JO)1.1 + .23Q .51 (.05) Brazil p 130.8 + o08Q .06 (.05) Argentina P 92.7 + .63Q .68 (.11) Source: See Tables 3 and #. Salter's data for the United States yield an i- of .36 for the 1923-1950 period. - 28 - unit gross margin to fal. together with falling labor costs which did no'v, occur in the three South American countries. A more substantial xplanation might be that factors uhich rake for grozth in production also make for growth in productivity. For lxaplei, an activity in which entrepre- neure ae alert ml cost-conscious is quite likely to be one in which both output md pmrductivity an gmwing. Expading industries attract better personnel and have better access to cr¢dit, etc. This argument, thouCh prima fLcie appealing, mac" furtr formnalization and empirical support.32/ It agroos with tha eNprience of tha United Kingdom and thG United Statos but cemot be sppliod in an unqualified ianor to Latin Anariean countries. Attention has alreaiW boon drma to Brosil's maverick st&tw (for xuhich no satiafact=7 o0plmation hw boon suggosted) and to tho rlo of mx - t in Argatina. In Snuth America, th3ofxv9 cnly °olbic scoK - at the moment - to fit the argument. Baplgmnt and Oa'ut Growth (Tale 7). Fbr both industria- :Eaed and developxing cowitris h tionship is quite signiftcant. 3a exeption io ArgantLna which ia consistsnt vith tha pzwiously a ggted wqlmation of tho ole of uadempbloymnt in that co=Ur. All regression coefficients cppar 1" rla t X zd lng of eiloynnt growth 3. An AtI22 P MMtq1C Eo ali ow umt,, migu es pictu¢e fro the jSaeading r zgy;ao~,, but sam findins og a hiwisetc natur dese emphasis. (a) ParhW th; aSt iEWM-t o soch findiag is that of the diMrawo batomm tlo Uzdtod Miugdm od Unidtd St&ats o- the one hand, nd tha thre South Ariom c ies 6 ths other, with respect to the relationshp beteu labor casts a 'br coswts The evidence seems con- vincing enough that B= tgi = St4ss and Utadg KingdoM sLl costs Loll together, Lo..the p=diwtivity og al inputs riav n or less togetor.2/ evien& is loss almw with roaet- to czitzl costs than ta materials costs, but em. oth i& no izSdLc:at= thati capital cost's 2/ C. F. Caer and B. L 15l11m,- i*nytSntt la Ta-mtiou London, Ox-ord University Pw,. 1958 d p s t in effective, although non-rigzaow f . is considorUb2a oth'r avidence alao, Soo. for exopl, 1fZ7' G4, AJlW,t. "IhzWUtlV Decisions in MAlrde"loped Qmta, v in tioE L . orfamic Growth, acuetoa, FN= Jersey, ?1uot= Uniwr&tty PrL. - d the various pE:prs in The Rate md.2zwt1m. < Inv%..t.Y Al "itt Richard R. INJIso (edi.tor),, Pri= ,, Mm. JermeT Prlc7 i,mrsi ty Press, 1962. 33/ Thais is eviaw a t th.s ni ±- o -.r bUrupean countries saE. taUl- IS tp Bt r cited .- th-0 goaces to Table 1. Table 7 Rel&tionship Between DRloynnt (E) nd Output (Q) in Selected Countries Couitr ruation _ TJiited Kingdom (1) E - 61.1 + .28Q .85 (.02) United Kingdom (2) E - 4.1 + .4i6Q .53 (.07) Oolonbla E - 105.2 + .23Q .51 (.05) Brazil E - 15.8 + .60Q .87 ( o5) Argentina E - 52.2 + .18Q .39 (.06) Sources See Tables 3 and 4. For the United States the rank correlation coefficient is .82 significent at tho 001 level for the Se-ter data. - 30 - rise with increasing leaor ,roductivitly. In the three Sout-h American countries, onlJy the latte: nmgatiwe ganralrization is supported by the equations. A poasiblo intGrprtation of the United Kingdom - UInited States results has alreedy boon suggsted, nme)y, that the sources of producti- vity grosth affect eaL' inputs morw or le9s simultaneously. If, as in Latin. Amarica, only labor prcdutivity risaes with no indication of a link bettaen it and the productivity of othor inputs, a differnt argument is nGoessary. Tao lines of inquiy amr suggested. In the first place there is the possibility that lbor productivity growth in the South American oountries is 8xplainod in t.ercD of iproved (not mor) cEpital that adds to labor productivity and of±f"te ew decline in capital productivity attributable to a rising capital/1'abor ratio. Since a lerge part of tVe physical capital awd;in thsGa aoeatris is lported, this would maan that OQir z2jor source of px:gntiAty grSo ew also importod. 47. In the ccchd plco (Sd not t of tho first place there is ttg poGoibility V;at VAn ooouoe env1ironmeant of Colsmbia, Brazil, ld Argntina is not ondhwive to gwanl inceases in productiv-ity. One ronson z this l in fcet be the cone for the entire economy was sug- gwted in Sactioa B ef Ctor II, as dtu to the particular typs of import gubtitution policioe folloic in tha comtries considernd there, Now anther source of diffi.lgy = te noted, w=4, that the physical capital tmt ia . (= that dzos prodco tho highor labor produztivity in oortoin cativitico) Lo aloo ualie1" to Latin Ameriem countries. The alien Zatlc not osay m=tuanc the omloy2nt, problem, but also imoses addi- t±sml dox& en c n oece that it sC mom difficulty in maeting (e.g., 2= anterial 1zDUtop okMcd 1intenanoe perforonce, otc.) than dono tho oc W In tLdat the epital WM constructed. This latter situa- tiam cda to to twmitzAa of the proeoss of adju,t ent to nrw processes, rLa tt=by cantributsto tho problea of raising the productivity of all igmt0o. In tbie mse teo cu5tnt tnek iposed by tho n= !talienR oc9iftl for the i ,orUg aoitry- is greetr than it is for the originating otry, t&ile (it oes safe to os) t2wir capacity to edjust is less (as rofolcted for eonm o in the rlative vsAati,ns shom' in Table 3). This 512M55tad bpo,asis am to te lnto4rpmt2atin of the sults thsrefore has two irportent iplioatiowo (1) that the thnre South Amrican countries mav had great diff1cult in eatablishing and meintaining an economic anvironmant comducive to produtivity gz&th, and (2) haey had varied suecoss in adopting imported technology without creating considerable dis- tortion, &ioh in tur ihped. poductivity grovtk. Met this inVirpreta- tion is of a heuriatic nat es d in no wse a logical consequence of oui' cqations io to be phasised.. lSo Attention als be called to the fact that r to of Table 5 (the relationship betot ot labor costa and productivity) for the three South Airican countries an8 Um below that for the United Kingdom (1) cwd the United States. Part of this differenee is doubtless due to the diff:or=0c in the ti$ ep , but not aLl of it, and w ar entitled to conclude fro this that the drveloping countries are less successf'ul in tomnferring the increased labor productlvit7 into loe3r labor costs than ca tha United States and the United Kingdom. And this, as pointed out earlir, handiaps the process of the kind of structural chan.ge considered in this ppor. - 31 - (b) The pictue for Argentina is a bit clearer tlan that for 3razil and Colou-bia. The close relationship beteen productivity and earnings per wortar indicates rathar convincingly a rather sevemly distorted labor market. The oquations in Tablas 6 (between producti,vity and output) and 7 (amployzent and output) both are consistent with an explanation of moasured productivity growth in termas siply of under- employment with little or no "puret productivity growth. Argentina with,, an older and possibly more sophisticated manufacturing sector, has also been more obviously a viotim of Qw kird of distortions rsferred to earlier than have the other countrios. 49. Itmayi2 o be noted, i passing, that ta rate ofgr tof capital fo.L ation hss been as high (or higher) in Argentina (investmant as averaged betwen 18 and 22 percent of kW over the last decade accord- ing to the national accounts) as in oth',r countries. This fact suggests that whatever prod.tetivity incraasing souce is tarried in new capital caanot affect t: 4iullifying effects that aria, from continued distortions of the kind summarized in Chapter II and analyzed in more det&1l here in Chapter !1. (c) The Brazilian picture is more clouded. The chief puzzle is the relationahip between productivity and output sbon in Table 6. The large (relative to the mean) varticwl Lnterccpt a=4 insignificant regression coefficient suggest strongly that measured productivity growth in not affected by changing levels of undoremployment, so too does the fact that r2 is .87, and the v.-T;al intercept in very low in the rela- tionship between ernploymant a&-d output. Theae equations plus those in Tables 4 and 5 suggest a rather pretty picture: a high rate of growth of labor productivity (about 2.8 percent per annmi eor the decade reviewed here, assuming that all the measured prodctivity growth associated with growth of output is duie to changing levels of undereployment), little underemployment, and a 1o>or market acceptably diaatorted. There is no doubt that Brazil has managed to avoid a r;zer of the difficulties which Argentina haa not, but stifll there is the point of sub-heading a above that no evidence exists as to other input costs falling along with labor costs. What it does seem acceptable to cmiclude, given the results of the equations, is that Brazil was - at least during ths 1950's - much more 3uccessful than Argentina in adapting to thb demnds created by the import substitution process.34/ This adaptation in th.e present context seems to be an important elemeiit in Brazil' s favorable showing relative to that of Argentina. (d) Colombia falls somowhat between Argentina and Brazil. Her equations in Tables 4 and 5 sugzest a reasonably efficient labor market. The equation in Table 6 suggests that some of the ocserved increasc in 3V/ Sor a recent survey of Brazil's industrializing experiences, see Werner Ba.r, Ind:tralIization and Economic Developnernt in Brazil, Hoimmfood, 1ll=nois, Ricaard D. Irvin, Inc. 1965. There is little doubt that the nicture naas chaged considerably since 1950. indeed sone tentative calculationz for the post-1959 years in Brazil suggest a :Jicturt vory iu=ch like that described for Argentina. - 32- productivity is an underempl.oyment phenomnon. The vertical intercept intthis equation is Just barely in excess of 100, uhich (as alrea&,y noted) suggests that the iacrarnnt in productivity over the period would have been slight had output not risen. Then in Table 7 the intercept slightly in excess of 100 i0r-8s that productivity in fs&ct decliled. Both these pieces of evidence suggot that increases in "pur" producti- vity were not groat over the 1953-63 dsetae. The bits and pieces of data on underutilization of capacity indicate that Colombian manufacturing did not suffer from this ma3lac to the extant that Argentina did, but there is evidence that Colonbie wo cwnparianeing incraeing difficulties in this rmopect in the early 1960180j5/ A r±ld nogactiv conclusion might be risked to the effoct that ths Col10Bo cquationw ;eQ not consistont with clear cut success of an industrialisati=a offortp tile at th:9 a= time not ring a disarray e ovidan c that in -gntina. A qiW8tion wall worth futriiar inquiry for both QOlozbia c>d BrOSi1l is that of whether their picture aY, dscribod by tbs varioug oquationg =uld be altered if data sare available to extend the perlod of nalvnio to 1966. D. This chater h8 att"mptod to G&W soiathing about sources of pro- ductivity growth and the proes of adjstment to increased productivity in Argentina, Brazil, and Colowbia. Yho stu%y of the sorios of regressions, thgh open to m=W quostienw, did su Seat soma geneeflizations about the aa%nro and soumo o8 tho dowlpmat difficulties that these countries face. A fial oeXpter m cotr to t a 0oy or tt^o by which these risults end thjse Gunerind in Mcptor 11 am rlatid to policy snttars. IV. SOE FOLIC DILMal Si. faig poper hegc bgaty ang t& O of outpat ecnot be sc+tis- aorily explainod in tozn o aroth C izputs as thago latter arm con- ¶antionally eamurd, and tet Inc esi pxcdutivity ias a Imort-ant part of the growth story in a2mt oal of tho c rich countries. It was further notd tht ts aa of d productivi ty ¶ usually in to= of i2roved skillst, teaAl change, increased heath standardB,, tet., and that now of tat ur of grouth is free but Eat be acquired by sen±Utuze of ins _tble xrSace. It seon useful, therefore, to thirh lIa tema of a genersZIa caital fbnmation moda. that meagizes the variou2 sauew of produtLv1ty grouth and that the creation eg these sourees requraes the = of iintibla rvzaurcw~ hen a series o* d±z±cal. atsA 3 reported on that suggested Ua p 1ral concluBins dth zoopet tot aseerel LaTin Amrican countries: (1) tAt productivity grawth in gnmral hs, sincee 199 or mo, been quite o for° 0 , Jobs B. m "Isrtar, Inwstx nt, and Growth: (b-lo bien Evezence Since 19%1" forthcming, in the collection of papers prsisant at tha Bllagio Canferece of the Ffrvad Developnent Adviso:: SO:@ . - 33 - Modecv, and below that of the 19h0's, and (2) that there is little evidence of the existence of an economic environment conducive to a high rate of productivity growth. Indeed a variety of policies common to most countries seem to impede the process of adaptation that uould facilitate the exploIta- tion of the possibi.tties for increasing productivity. 53. Once it is acoepted that increases in productivity are important .n understanding growth and structural change and that they do not fall freely from the heavens on all sectors in all countries with equal inten- .*i Jy, vhne Lt is evident that there are major policy issues involved. A fuil scale review is not possible here, but soma of the more general (and obvious) points may be noted. 1. Size of Investments in Infrastructure 54. The most ev-ident point to make is that the invastment allocation problem is extended to include the range of activities a'fecting producti- vity. One must .ask about the productivity of expenditlures on educition or on cechnical research or on health as well as con ateel mills and chemi- cal plants. Rates of capital formation (conventionally defined) appear generally acceptable in Latin America, and one possible explanation of the difficulties observed in these countries is the simple point that "too much" has been given to the building of plant and equipment and "too little" has beeni left for productivity-increasing activity. 2. Efficiency of Investments :.i:nfrastructure 55. The previous point must be qualified to some extent by noting that the efficiency of resources In productivity-increasing activities may be quito low. Thus it may be that trained labor will result in greater labor productivity than untrained labor but it mny also be that the quality of the educa'ional 3stablishment is such that it is unable to teach ihat students need to learn. For ex &le, students attend school to learn to read rand write but do not learn to do this. The difficulty here then is the quality of the educational establishment, not the nuirbers sent through the mill, and the allocation problem is that of i3proving this quality. This latter is, of course, muc. more difficult than siDply constructing more school buildings. There is some evidence that the education bottleneck is of this sort in a number of the Latin American countries. 3. Difficu.±ties Ln T9olicy Changes 56. HŽrqphasij was placed in Chapters I and II on the hypothesis that zhe particular nature of the development strategy followed in Latin America has had adverse effects on both productivity growth and adjustments thereto. coa3mic policy springs from a variety of considerat4ons, and ene ca-nnot cj.. ;policy is wrong simply because iT hampers productivity growth - granting that it in fact does irn a number of countries of Lat.a America. It does 6oeen ciear, however, that, sxamining the effects of specific policias on ro.4utirity g- th - given the demonstrated imnortance of the latter - ' C. fruitful approach. Suppose one mre conv-L iced that the relationships IMIe=n productivity grot4h and devolopant policy summarized in this " were valid. Could oZ then conclude tat all the countrios had to ,Q was to elininate some of their mom damaging policies, ard productivity 1owth mould jurp upward? it is probably correct to argue that correcting. p ,Cy is not a completely free altpraative. Fbr exaMle, sappose cone 4g$~ that the tariff structura was "too hig" (or the exchang rate wrong, It is sureJy correct in yrq instances that r6duction of tariffs (w va^,luation, etc.) c. impo temporary cost an the eociety. This qwp3kw ac 'ion may the* be roceos;d but poitpoued indinitolv _aca.se e % short-run costs incurred in its iplementatiLn. Then aieG may be R used as a ve.Mcle for carrying out policy cAzangea, i.e., in i0 w 14e costs iLnvolved in shifting from a "wrong4' to a lrightV policy. 5,0N This point- has a wirieit o;?l impl.cattons, but only on will be ,2k*~ u that hai"t to do xtA internatimenl grants and loans. One q,t Is Joplications of itea (1) above is that ait (mDaning all kinds of R,4aAiq qqpital flaus) could profitably bo used to support productivity- 1DqqQa?i3W-e ^+tivities. Now added is the point that aid can also facilitate (/Q;p. bWt) the implemeatation of major policy ckanges. Indeed foreign aid 4W4 fhuZ conBtruction physical capital or simply to roli&ve a balance of ? Qrisis qan, in fact, iped% tiha iplemEntation of corrective policy The incoma froi the physical capitAl becomes a substitute for RB yiRAWty growtht And reaouroe 2X9 provided to support policies (or i@Jt,blJ0Q,jis practices, aod ea on) An'idca2, to. productivity growth but UMqUIA R;l painful to chaWne. Tt iD a ceomlicated issie but the central jjk. Qt talvW>ce in thb Ir?sFatt oon>^t ia &ixple enough: ceortain policies = Zkg"tRJ4A1t4Tity grzsb. ednxm7 enough that an alternative policy is - kgQR eE,. le~eEte4.t a desirable. The poJiW. chanh e izpo me certain Q CQSt4- (e.s.,. caital 111 b oli obEa.tbsl, distribution Q4' tJQ9BSS zSt, LA, etc.); 45 RR;L a i1,os1ia2t otbar,. more intractable social to'. Ai&s tw Erq lav a iSa higha pPreftctivio if it is tff AQ to itate policy chm%lm by on t>H4 tbgi- eoozmic coatse involved ig b"f cho . TbB dancer ia) ho'gvaz'M tbia& anuh,,. policqt changes often b$sg difficult and trouta somm tsitLa& = inaitutional chauges,. aid may z* iAwA to ride over a balmuct ofFj RWBmYr BrblF mitiout Q ibEknatirg ltjj w4lxmate source. To rejqaA tbb kee poixvt hwzw2 policy, ahange (e .g., W a tariff sche4jae)' cam, lijpoDo short)w. Co i4-ile yielding long- a, Aid to finance theaw shocatu*t , tn: be2 ai productive a ok *kqh resources. b t 1U-rc2-1.DvAt4QJ; via*~r .bii*tdt1 Qmoiu. Much of the. arguarA tipa pqpr- h2 Is t, conce-usiouzta that {b^9^23MMuio of imsouxcea ; adrare& e*Ztqeiita o= Pro4notd vit. growth and. t* s, *Atlye markets crx.ted dUf1toul.t,.ar ft the spreading out. over the gMy soJP. tf rewards of. produpt4vity grawthn.. Roewrtly a nw.b.r of economists var:2& tbaj. eliminat.4iQ o. tbe; rAallocation or risoureee can add little - 35 - to che capacity of an econony.36/ These studies have, however, looked at o.ily the static consequences ormisallocation. If misallocaticer does dampen productivity growth, then the allocation question may well so zrucial to the achievement of sustained growth. This conclusion, one maW r.phasize, is not equivalent to saying that if a countrf solves its alloca- t.ion problems (and prevents inflation) it will, thereb:, enjoy sustainod ,rowtLh. 31/ vhis material is sunmarized in Harrey Leibenstein, "A1.ocation Efficiency vs 'X-Efficiency'," American Economic Reviewm June 1966. APPFhEDIX STCTIURAL saNYE AND P1O)UCrIVITY IN LATIN AMERICA A Survey of Variation 1. Table 3 /37 contains estimates of the relative variatioit (the standard deviation of & series divided by its mnaan) of the series to .iich attention wil. be directed. It is evident at once that for each country the variation in gross ou°put exceeds that of a11 other serias. Also generally high is the variation in employmnt, though Argentina appears to be an exception in this raspect. More iNpwrtantly e&rnings per worker ir.dicate deoidedly less variation than either output or employmant (for Argentina variation iv eamings and employmnt is about equal). It should, howevar, be no4.3d that t,œ 3xlative variation of earnings in the three Latin American countries is markedly higher than it is in the Un:ted States and the United Kingdom. Of equal zelevance to the present argument and to other aspects of productivity analysis is the observed variation in produc- tivity growth. If tha variation in this series were quite small, this fact ouLld ; ggest that the sources of produotiviL'y growth (uhate'!er they are) act more or le-is equally on all sectors. Consiterable variation, among activities on the othbr hnd, suggests that proluctivity-increasing forces are not equally strong among sectors ol. that zrjaction to such forcee varies from activity to activity. Indeed, it is this variation in produrtivity grouth that mxakeL tho kind of analysis undert.aken hre useful. 2. The final three rows in Table 3 have to do with costs per unit of output for t-he three identified inputs that make up gross output. Unit labor cost (equals earnings per worker divided by productivity) show a variation that is due to that of its two components already mntioned. Unit material input cost (value of material input divided by gross output) is an attempt to measur the ext-nt to which this component of total inputs changes as output changes. Since all sectors use certain commn inputs - e.g. fuel - the variatior' among activities of this item is expected to be relatively small. The gross margin is defined as total gross output minus the wage bill minua material inputs (or va'ue added mn1sus the wage bill), and this divided by output is gross margir cost per unit of output. In a very rough and ready way it can be thought oi as a measure of the cost of capJ tal per unit of output. Variations of this item differ markedly among the countries shown. This figure is the second highest in the United Kingdom and Argentine colunmns, the second lowest for Brazil, and the third lowest for Colomibia. It may be specifically noted that unit materials costs and uniL gross margins are not measures of productivity. In both there is a price element that cannot be factored out. 37/ For co-avenience, Tables 3-7 already shown in the main text are roproduced here, with identical numberirng. Appendix - Page 2 - TabIw Relatin7 l=iati= ga Sbleted &,is (Standard deviation dirided by muan) 8 L 3ttSi M31'd Ki iol _________ 3anzi1 ret qSS; output 5500 &?2 T.8 53o63 61.27 Muayat 3tv gtl _l5 51.6a 35 35 l=*AP U. 2a7> A 22.29 15.17 35. 1? Efctivity 30.& 35.T7 571 20;.69 4lC.07 ouklabor cost Z7.3 AA 3445. .36 17.68 &gmmaterial cost S 13e2F 9.15 32.29 1=t. groass mnrg.n JZ ne22.75 U7oT4 5io77 1.: tbo W.e St&tcw t!di t figursa taken from the SUr > cftmvaa- iz fwut 31L nw CozmES B,3, _ - t e for calculations obtained Sw= dwauzla 3wtt1iUw cI£ Bra1U United Nations proc a Xt i ME R 1 atacl A=w==VttM2MwAitc fMai- Anuario Ebtatia.co cbaaIndas and uzrst I have collOc ad? 5= zo- fcia. w stica1 eencies in these aouatriu. Tim NPri: d 1 _ I99J 28 activities Units& StAdMa EI-197 27 activities Colon*a 195-1963 20 activities Brazil ISW-19957 18 wctivCties ArgentIna 19-.l962 15 activities Appendix - Page 3 - 3. The task now is to tr; to explain some of this observed variation, es?ecirlly that associated with productivity and to explore a few consequea- ces of these explanatio.s for *he developmwnt process in Latin Amarica. The Sarnings - Productivity Relation3hip 'I'lie most obvious and ost convenient place t4o begin is with the rdlat.ionsl:)is between earning per worker and the rat'- of growth of output per '¶ork-er. A3 noted above thls relationship may suggest something about the nature of the labor market and possibly something about the sources of orcduct.ivity growth. This relationship for the countries under considera- tion ns presented in Table 4. 5. The only maverick in the table is Argentina. Zor all the other co.-itri.es variation in e--nings per worker seems to bear no relationship at all. o variation in Productivity growth. The higher intercept vAlues for the three La`in American countries tihar. those for the United Ki-gdom and the Uni 'od Ste.tes reflect the fact that money wage rates have risen more rapidly in Tatin America than in the United Kingdom and the United States. Ihis in tu;-, reflects the inflation 3n these countries. 6. Atten.tion was cal-led earlier to the fact that the calculations shown in Table 3 indicate that t'he relative varzation in earnings per worker is greater in each of tthe three Latin American countries than in the United Kingdom and the United States. . result is not sarprising, although the clifferences in this variation between thie Latin American countries and the United Kingdom and the United States is probably less for the manufacturing sector than would be the case if the whole econonV were considered. Part of the explanation of this difference is to be found in the general charac- t> ristico uf the labor sar-ets - in South America _.ess mobility, less informa- tLon abcut wage rates end *,ob opportunities, less spill-over effects from one sector tc another, i,.;' variation in the extent to which finrs confor-n to labor i&ws derinixii socia) security payments and other non-wage contribu- tions, etc .33/ Thus the J oor market, like almost all markets, functions less we.ll inthe Latin tmerican (and most other developing) countries than in the more advanced economies, and this fact is reflectec in the grea'er -.rarLatioa in earnings per worker shown in Table 3. For essentially similar reason,;. one migli have expected that relationship between earnings and pro- ductivlt; wot2oul.i have been so ,:Lwhat higher in all the Latin American ccuntries t>: S: t.he i:r.: i;l 'v.Tocoipute an index of mateHiMi productivity and capital productivity in the s5ae mariner that labor productivity has been copqiiled would be ..cst desirab'Le. To do this, however, in the present exercise was not pos;sibla as p.lca data suitable for deflating the two series were not roadily at hand. Appendix - Page 6 - Table 5 Relationahip Between Unit Labor Co¶t (ULC) and Productivity (P) for Selected Countries -2 CountyX Equation r United Kingdom (1) uLC 305.3 - .82P .82 (.07) United Kingdom (2) ULC 236.8 - .8isP .63 (.12) Colombia ULC - 41 o.3 - .70P .54 (.14) Brazil ULC 910.7 - 2.58P .? (.62) Argeatina uLC - 1306.4- .92P .08 (.61) SouSe: See Tables 3 and 4. Thr2 for the United States is .80 uing the Salter data. Adpczddix - ?age 7 - toQCtthr WLth the consequent reduction in price, and second, if the :)o:,- labor costs do not move with the labor costs, can we interiprut thi- to mean that labor costs fall (rise) because the other costs rise (fall) - i.e., because of a substitution effect - or can one find reason to conc2tdc that other forces are producing the changed labor productivity. 10. Consider first Salter's findings for the U'nited Kingdom.. One of the more striking results of Salter's computations is the finding for the 1926-1950 period of an r2 of .61 for the relationship between unit materials cost and labor productivity. The relationship is, of course, nez-ative.40/ For the shorter period the relationship is less iv)ressive (r20 .21lS but still suggests that even over this shorter period there appears enough evidence to Justify consideration of the view that a cormon source seems to act on both labor and materials costs. One can say then rather confidently that these results do not support the notion that the rising labvr productivity can be explained in terms o' an increasing use of (e.g.) fuels or a more extravagant use of raw materials.41/ A co-sres- ponding (positive) relationship is found between ULC and Uli7. 11. Suppose for a moment that unit gross margIn costs (LGITC) were an acceptable measure of the input of capital services, that might one expect as to tha relationship between UG14C and labor productivity? The simTplest (and oldest) argupsnt is that a strong positive relationship should exist between the two series indicating that capital was being substituted for labor and thereby pushing up the productivity of labor. Simil.arly, one would expect a significant negative relationship between UGMC and UILC. Salter's da a for 1926-50 for the IJnited Kingdom show the oppos'te sign to that indicated by these expec'.'ations, and in each equation about 12 percent of the variation of th3 dependent variable is explained. For the shorter period the r Is are negligible in both equations. Given the resi- dual and hence nebulous character of UGIC, perhaps the least dan4gerous interpretation of the resultz is the nsgative one that there is no evidence here to support the assumption that labor productivity rose because of an increased use of capital in the productive process. The stronger conclu- sioIi to the effect that UG1{C fell as labor Droductivity rose (and as U:LC fell.) is consJstent with the evidence for the 19o26-)3 period, but not for the shorter, second period. 12. (Given the preceding results, it follows that there is a negative and significant relationship between t4e rate of growth of labor producti- vity and price. Fbr the two periods P' is .77 and .27 respectively. 1O! The corplete equation is UIMIC - 334&.° - .78P where UM1C is unit materials cost. Data are i;ot available for the United States. 4'/ Price of materials input could move in such a way as to contradict the argurent in the text, but this appears so improbable that thc possibility may be ignored here. Appendix - Page 8 - 13. For the three Latin Amarican countries the results are less clear. For none of the three is ther a statistically meaningful rela- tionship among the riables Just reviewed for the United KIingdon. The calculated regression btweRen unit materials cost and unit labor cost or between the forar and labor productivity show 8s of virtually zei'o in all cases. The a8 holds for regrassions betteen UGI4C and ULC and pro- ductivity. The nearest exception is for Colonbia i*here the ragression UMC - 237.9o + .26P 9 .12 (.140 is not inconsistent with ttm notion that labor productivity rises, at least to soma extent, because of a substitution effect. The results then for Colombia, Brazil, and Argentina are essentially the sase as those for th3 United Kingdoia with respect to UGMC and labor productivity and labor OOWt8. Tho Latin Amrioan co=trioa show a quite different situation with rmopect to nbariale eoto ad the labor -ariables. Consequently, nothing op.cific cm be dedu6ed about tho relationship betwen labor productivity and prices, althovgh we do have enough information to conclude that in general this ralationship should be quite izsignificant. Price indices, sowevera rot such that one can co1pute a productivity-price regression directlyo . ESBafora oosicring eaO iM1ioation8 of these riwults it is help- frl to look at t additional ste of ralatlonahips.L2/ ),/ It was not poasible to obtain data and carry out the formal conputa- tions for the United States with reapect to the relationahip between productivity grwth and capital and materials costs. Preliminary calculations, howver, indicate a picture virtually identical to that of the United Kingdom (1). With Kendrick's data for 33 manufacturing activities a rank oorr.lation of -.59 is found between "total produc- tivit3M ad terials cost per unit of output for the period 1899 to 1953 and ooeffioients only slightly lower for sub-periods within this long period. (A .0M significance l l requires a rzank correlation of .43.) The am general resat holds if one uses orly labor (inst3ad of labor and capital) productivit7. Sinilarly, a significant negative rank correlation coefficient is found between labor productivity growth and "'unit value" (i.e., price) and total productivity growth and price. Thus the evidence seern quite convincing,, depsaite the crudity of the rank ccrrelation coefficient, that producti-eity of all inputs moves together, and the rate of growth of productivity of labor cannot there- fore be explained in terms of a substitution argument in the United States and the United Kingdom. See the Kendrick volume previously cited,, especially Capter 7. Appendix - Pa6e 9 - Produc tlvit_y and Output 15. Table 6 proeants thX regrassions between the indices of produc- tivity growth znd those of output. In this table 3razil is the marerick case, as all the other countries show a significant relationship between the two variables. The first question to ask is simply why should one cxpect a relationship of any kind betueen productivity growth and output growth. There are sereral possible explanations. 16. The most obvious - and least interesting - explanation is in terms of cha.nges in the degree of the utilization of the employed labor. If al1 labor were not used full time all the time then labor's average product would vary with output simply because output could increase from a position of underutilization with no increase in employment. This would mean that soms part of the obser7ed increases in productivity did not repre- sent ',true'! increases in laborts productive capacity. A clue as to the importance of this effect is provided by the vertical intercept values. These values indicate what the productivity index would have been on the average had output not increased at all. An intercept value below 100 is most easily understood therefore in terms of an underutilization ar7--nt, since it Is difficult to believe that "pure" labor productivity - i.e., labor productivity corrected for any underutilization effect - declined over a period as long as a decade or so. 17. For both the United Kingdom (2) and Argentina the vertical inter- cept is less than 100, and what we know from other sources about these two countries in the period covered is consistent with an underemployment inter- pretation of these equations. L. the United Kingdom 1963 output suffered from t}e aftermath of the balance of paymants induced restrictive policies * 1962 and the effecta of some bad weather. In Argentina underutilization of both capital and labor has long been a major problem, although specific data on the changing degree of underemploymant over the period covered by the data axe not available, there is little doubt that it was substantially greater in most activities in 1962 thar. it wa in 1950, and that it was greater in Argentina than in Brazil and Colombia.L3/. The argument further requires that employment not move exactly with output (this is what under- emplo3ment in contrast to unenployment means) and as shown below output and employment were less cIo sely associated in Argentina thar. in the other cotuitries under reviet. The regression coefficients for Argentina and United Kingdom (2) ir. Table 6 ame substantially higher than the other :e.* gression coefficier.cs. Argentina's .63 especially is decidedly higher than the others. The regression coefficient, of course, indicates the exte:.t to whichl productivity inc.eases with an increase in output. A larger coefficient for Ar-gentina therefore is also consistent with the notion that the relation- ship betwecn productivity and output in this country is due to changes in 3Jl/ :3one data on underutilization of labor and canital in Latin Amerinc are found in the statistical arnex to ECA's Process of Industrlal - zation of Latin America, United Nations, 1966. Appendix - Page 10 - Tabie 6 RalationBhip Between Productivity (P) and Outpflut (Q) Lor Selected Countries Countr on 2 United Kingdom (1) P f T3 + .23Q .65 ; .03) United Kir- (2, E ! 72;.- 43I1Q D46 (.08) oo1bmbia - E=10...- .23Q .5 (.Q5) Bra.i. & w .o8Q .o6 (.OS) Argentine F v .& .68 Source: Sa - } 1& 3&1z4 w data ftr th. United States y '. m Pf or .36 fobr. tb2 1923-1950 period. Appendix - Page 11 - the degree of underemployment. A similar argument would apply to United Kingdonm (2). All these bits and pieces of information pl.us the equation in Table 6 suggest rather strongly that a major part of the average observed incre se in productivity in Argentina is due to changes in the degree of underemploymant. nhe sam seems true of United Kingdom (2). No doubt this argument is applicable in some degree to Colombia and Brazil as well, but obviously aot sufficient so to provduce a vertical intercept be. ow 100 or to produce a very high regression coefficient. 18. The observed relationship between productivity growth may also be d-:o to economies of scale..4/ One would expect, however, if the scale effect were predominant, it would be reflected in all inputs, not just labor. As already noted non-labor costs do not move writh labor costs for the three South American countries, and less clearly so with United Kingdom (2) than with United Kingdom (1). The scale effect thAn seems less apparent than one would expect if it were carrying substantial weight in the explanation of P. 19. A riore substantia. .explanation of the relationship between growth of productivity and of output has to do with the argum4nt developed above that productivity growth is not only an important source of growth but factors that rEakc for rap,d growth of output also rake for rapid growth of producti- vity. For example, an activity in which enterpreneurs are alert and cost- conscious is litk-ly to be one in which both output and productivity are -rowing.)&/ In a similar vein, it is probably correct to azsum3 that acti- vities u-ch are achieving high rates of growth of output attract better tralned labor and have better access to finance than do the slower growiag acLtvities. This general arximent is appealing in a variety of ways, but it obviousLy needs further fonsialization and further empirical support. 'some evidence ias bcen cited above (e.g., the absence of evidence that incrementi of labor productivity are due to the substitution of other inputs for labor, the absence of a relationship between the rate of growth of in- puts and the rate of growth of productivity, etc.), and considerable additional evidence is ava.ilable.L6/ It is at once evident, however, that the ar.-ument cunnot be applied in an unqualified manner to the Lat4;i Amezican countries. Attentioi has already been called to the role of undererploym-7vt in the case Of Arlt;etiiia arid to Brazil's mavarick status in Table 6. O.i.ly Colombia in .outh .A-irica -;eemn - at the moment - to fit the argument, while the evidence for the ?nittcd Kingdom and the United States is considerably more consistent. s Wful. 1- 'ook at one fioal set of relationships. LiLK owe *rL: ,; argue t.at -ne scale effect was simply another side of the u:idc riloyme'it e;fect, but it is rarely so considered. ;,/ T&...c. c;re sit.- of ev.dence on this point. See, for example, Choice o: 2e_ :ob-:-is ic.n o L-C-atin American Textile Thdustry, Economic COi:'.'ko33011 'Or ,.atin AM.-rica (mireographed) 1966. This argument does ot' course, &ythin- about the suitability of the more Iv':tn.ed "roIolg from 'the standpoint of the entire econonm. . . . / ,,, *, Appendix - Pace 12 - dlo;02nt ond Ou-tout *2rowth iLn Table 7 regression equpt±ons 'Detween indices of enploymenI. and output are show.n.. Each equation represents a siiLificant relation3hi,), and as indicated in the note to the table, there iui doubtless a si,ni.fircazit ralatio,nsh.ip between the two variables lor the United States as well. Thlat such, a relationship exists is of course not surprisi:g. For there to be no relat'onship would man that labor productivity grew at the same rate as output, and this is quite unlikely. The equations in this table also t%veal some of the points made in thI discussion of earlier tablcs. The l)a regression coefficient and low F for Argentina are consistent with the role of underenployment in this cojpntry noted earlier. Simi.larly, the YIhgh *&ession coefficient and high 4 for Brazil is to be expectad, given tls 14 correlation batwzi productivity and output for this country in To 6. The higher regression coefficient for United Kingdom (2) than ifo 1 nited KingdoQ (1) is also consistent with previous argument. A puzzle damTs aypear, howeve;, with respect to the vertical intercept for Colorb.a. The 105.2 indicates that e.4ploymant would ha:e risen even if output had not, eLd this is difficult to accept. Taken literally it would mean that labo- pTodutivity declined over the period considered, and while this result ight be made consistent with other evidence, extended argument would be required. Sone note, hauever, will be taken of this possibility in a later discuission. ::t may be noted in pasing that all regression coefficients apear rsInall" - even that for Brazil. This result reflects the well-known fact that esiloymant lags markedly behind manufactuaing output in all counries.Q Appand.ix - Page 13 - Table 7 Rel.ati2aE!Lip Between &ipl1oymeat, CE) anid output (Q) in Selectad Countries Country Euationr United Kingdom (1) E n 61.1 + .28Q .85 (.02) i%tiWd Kingdom (2) E 414.1 + .146Q .53 (.07) Colombia E - 105.2+ .23Q .51 (.05,) Brazil E - 15.8 + .60Q .87 (.05) Ar, i.ttina E v 52.2 + o18Q .39 (.06) Sourze: Sae Tablos 3 and 4. For the United States the rank co.rre1ation coefficient is .82 significant at the .01 lavel for the Salter data.