Warld Bank Reprint Serieo: Number 254

Ceashon Feder

On Exports d Econo

Reprinted with permission from journal of L'ev~lopment Economics, vol. 12 (1982), pp. 59-73.

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

World Bank Reprints

No. 219.

No. 220.

Nc. 221.

No. ,222.

No. 223.

No. 224.

No. 225.

No. 226.

No. 227.

No. 228.

No. 229.

No. 230. No. 231.

No. 233.

No. 234.

No. 235.

?do. 236.

Trent Rertrand and Lyn Squire, "The Relevance of the Dual Eccpnsmy Model: A Case Study of Thailand," Oxford Economic Papc9rs Cershon Feder and Knud Ross, "Risk Assessments and Risk Premiums in the Eurodollar Market," Iourr~al of Financr Bela Ralassa, "Policy Responses to External Shocks in Selected Latin American Countries," Quarterly Review of Eronornics and Busincss Choong Yong Ahn, Inderjit Singh, and Lyn Squire, "A Model of an Agricultural Household in a Mul ti-crop Economy: The Case of Ksre;," Rre~irzc~ o f Economics arid Statistics Emmanuel Jimenez, "The Economics of Self-help Housing: Theory and Some Evidence from a Developing Country," Iournal of Urban Ecntiotnics Thawat Watanatada and Clell G. Harral, "Determination of Economi- cally Balanced Highway Expenditure Programs under Budget Con- straints: A Practical Approach," Transport Research for Social arid Ecotiornic Progress George Psacharopoulos, "The Economics of Higher Education in Developing Countries," C,vnparativc Education Review Katrine Anderson Saito and Delano P. Villanueva, "Transaction Costs of Credit to the Small-scale Sector in the Philippines," Ecorzoniic Dez~eloymcnt and CulturaI Changc Johannes F . Linn, "The Costs of Urbanization in Developing Coun- tries," Econoniic Dez~cloyment and Cultural C h n n g ~ Guy P. Pfeffermann, "Latin America and the Caribbean: Economic Performance and Policies," Southzc?estern Rcvieu~ of Management and Ecoilolrlics Avishay Braverman and Joseph E. Stiglitz, "Sharecropping and the Interlinking of Agrarian Markets," A~iicrican Ecoriotriic Rez~iew Abdun Noor, "Managing Adult Literacy Training," Prospects Bela Balas.;a, "Shifting Patterns of World Trade and Competition," Grozc7th rind Entreyrcrlenrship: Opportunities arid Challcil,gcs in a Clzangi~~cg World Johannes Bisschop, Wilfred Candler, John W. Duloy, and Gerald T. O'Mara, "The Indus Basin Model: A Special Application of Two-Level Linear Programming," Mt~thotlafical Prc~granrntin~ S t ltdy

Keith Bradley and Alan Gel$, "Motivation and Control in the Mondragon Experiment," and "The Replication and Sustainabilitv of fhc Mondragon Experiment," Brifislt lorarnal of Iitdlrsfrlnl Rclnt io~~s Gary P. Kutcher and Roger I?. Norton, "Operations Research Methods in Agricultural I'c llicy Analysis," Eliro~~cntl Iournal of O~~crntiotlal K rsi~rrrcit Bela I3alassa, "Econo1r,ic Reform in China," Bmncrl Naziorlalr dcl Lrzs)c~ro Qlrrtrf crly Ri9z1li.zc1 S. van Wijnbergen, "Stagflationary EHects of Monetary Stabilization I'olicies: A Quantitative Analysis of South Korea," ](~rtrjlrzl of Dcz.cloy- tnc3r t t I:i-clrrcln~rc-t;

Journal of ilekelopn~cnt Eeotromrc\ 12 ( l ( I )H2) 59 73 Norrtl-?.litlland P~lblrshrng C'ompilnj

ON EXPORTS AND ECOIL'OM16 GROWTH"

The p i i p r analy\e\ the \t)urce\ of grouth :n the prlcld 19M 1973 for a group of seml- ~ndustriairzed le%\ developd cc)urltrlec 4n analjtr~dl f rdme~ork is develowd lncorporalrng Ihe po\\~htllr) thd: mdrgin;ii factor prociuct~\tt~e\ dre not eyu'il rn the expclrt nnd rton-export gctors of r l ~ e ccononry. kcont~metr~c analys~\ utillrlng !hi\ f r ame~ork rndrcate\ that mdrglnal factor p r~ \d t i c~ i~ i t i r \ 'ire \~gntficantl! h~gher In the export sector The dlirerence seems to der~be. in part. fron~ rnter-~ect(trdi hencfic~al ertterndltt~e\ gencrdted bc the crporl 5rctor The cr,nc!uslon 1s therefore thni prouth Ldn t-e acnerated 11Oi on!? bq ~nzrensc\ In the dggrepdte feteis of jakor and cdprtdt. hut d1\0 h j the reailocdt~on of erlsting re\c>urce\ from the le\\ eficient non-export \ector rct the higher producrittty expctr! sector

The relation k e t ~ e e n export performance and economic growth has been a 5uhject of con3lderable interest to deketopment econornjsts in recent years. E-mptrlcal cjhiertatlons across countrta5 fended to demonstrate that deteloplnp countrie\ vrlth a favorable export growth record tended to enjoy higher rater of g r o ~ f h of narlonal income. Obviouslj, srnce exports are a conlponent of aggregate okitput. one if;ould expect a positive association in termi of the correlat~on coeffiiclent [Krakis (197011. But several ernptrical studnes demonstrate thcrt exports contr~hute to GE3P growth more than just fkc change rn the bolume of exports [Balac;$a (1978). Heller and Porter ( 137111, Itiichdelj 4 i 9 7 ~ M~chalopoulos a r~d J a j ( 19731, 7 yler ( 198 1 )J.

Crplar~dtion\ for the\e ob\c.rcatrons ucrc dl\cus\ed b j a n u m k r of cconc3m1\t$, h~phllghtlng tarloui benefictal a s v e t \ of ekpc~rts. such as greater capLsc'l0 bitlllrirtlc3n. ccunnmie\ of scale. incentltes for technological rmprncernentr and efifiic~ent management due fa C O M I S ~ ; " ~ i t l ~ e pressures abroad [see I?ialCt\\,i ( 19781. Kecaing 4 1967, 19691, Krueger ( 1980). and Bhagwati and Srrn~aa(tar~ (1W981j The rmpltc;l?ron of t!sc\e discussions 1s that there &re \uhstcintlal difTerencei bervrt.~.~ mtzrglnal factor productiiities in export- ortented and non-export-orier11cd ~ndurtriel. ruck that the former ha\ e higher

'I he \ ~ P H \ cxpresecl In thl% pdpr &re rhtlre of the author 0 1 i l ~ dnd cannot he faken 1 0 represent fkc: \re%\ t ~ f the Wor ld Hdnk The author krsefiretl from helpful d~rcbiaslons w~rh R R ~ l d 4 4 ~ i h C'he,lcr). 1 I>~Melo, I I9cWulf. I1 Keeeing. G Pur\el.ll. "iohlnscjn. %I Fyrybiiin snd 11 Wt ir

trO Ci. Frder, On rrporr.5 and cls.onrrmii. arowrh

kctor productivity. Ill fc)llows, then, that countries which have adopted policics less biased against exports benefited from closer-to-optimal resource allocation a l~d higher grow? h.

The present paper develops an analytical framework for the quantitativr, assessment of factor productivity differentials &tween exports and non- exports using aggregate data. This framework is utilized in an empirical study of sources of growth in a sample of semi-industrialized less develowd countries within the decade 1964-1973. The analysis allows an estimate of the sectoral marginal productivities. Furthermore, the extent of inter-seetoral beneficial externalities generated by exports can be spcifically identified. The results highlight the role of export performance in explaining the growth record of the sample countries.

2. Framework of analysis

The analysis in this paper adopts a supply side description of changes in aggregate output. In so doing, i t follows a practice widely used In the empirical study of sources of growth [see, e.g., Hagen and Wawrylyshyn (1969) and Robinson (1971)l. Within such a framework, where aggregate growth is related to changes in capital and labor through an underlying production function, the studies of Balassa (1978), Chenery et al. (1970). Michalopoulos and Jay ( 1 973) and Tyler ( 198 1 ) have includeci an indicator of export performance among the variables explaining g rowth .qn the following, a framework is developed whicli provides a formal rationalization for the incorporation of export variables in the sources of growth equation. Furthermore, starting from the sectoral production functions, a p ropr specification of export variables is indicated and a non-conventional interpretation of parameters is implied.

Since the present analysis focuses on the potential non-optimality of resource allocation between export and non-export sectors, the economy is viewed as if it consists of two distinct sectors: one producing export goods, and the other producillg for the domestic markets2 Instead of an aggregate national production function, each of the two sectors' output is a f~nction of the factors allocated to the sector. In addition, the output of thc non-export sector is devndent on the volume of exports produced. This formulation represent. the kneficlal effects of exports on other sectors [Keesing (1963. p. 31 1, 1978, pp. 4, 5 ) ] , such as the development of efficient and internationally

'Chenerj el 81. represent export? prfo~wance bj export prr3wth multlpl~ed hy expjrt \hare. the other works Include exprtc groath

'Clearly. this IS an abstractton, as many firm? are prtduclng for hoth domc\rrc and external market5 I t may be argued that eken then, !he dome%tfcd)ly mayketed output of suck firms I \

characterized by the same fiictor productl\rtie$ which characterize exptrts tdit?we\rr. to the extent that growth of exports repre\ent\ a good i8pprox1matlo-1 of the change4 in the bolunre of product~on of such firm?, the re\ult% are \ t i l l ~dllcf

csmvtitive management. the irnlroduction of improved productrrtn techniques, training of higher qarality labor, steadier flow of ~mpclrtcd ~npnls, etc. The* eflects are referred tc:, as externa\itics, sir~cc they are riot reflected in

market prices. These externalities are incorporated in the formulation tPt.lcrw:

N --- F( K ,, L,, X) , ( 1 )

X = ( ; ( K ,, Id , ) , where 2 )

N --- n ~ n - e x ~ r t s , X =exports, K , , K , =. res~c t lve sector capital stocks, L,, L, = resgective siactor labor forces.

Since data regarding sectoral allocations sf primary prodtaction factor5 are not readily available, a spcification is required which ~ 1 1 1 allow estimates of sectorai marginal productivities using aggregate (national) data. This is accomplished as follows:

Suppse that the ratio of respctive marginal factor productivities In the two wctors deviates from urllty by a f a~ to i d, i.e.,

where the slibscripts denote partial derivatives. In the absence of externallines, and for a given set of prices. a srtuatlon

where 6-0 would r-eflect ill1 allocation of resources vilhlch malrrrnircs national output. However, duc ta a nurnkr of reasons. marginal factor prductivitiea are likely to k lower an the non-export G G C ~ O B (I.e., 0 ;rOi.,Onc important reason is the more ~ ~ m ~ t l t b v t " env~ronment In which export- oriented firms operate. Competttion induces innovatic-enc\r, izdaptabtlit!, eficient rr,anagennent of firms' resources, etc Another reason for detralron\ ktween setoral marginal factcrr produ;t~vrttes are varioklr r e g u l a f r o ~ ~ ~ and constraints such as crdi t arld foreign exchange rationgng [Hlalassa (197711 The higher prcefved uncertainty assmiated with export enterprlsc5 ma! hi. another factor contributing to deviatiorns in marginal producai'ilt! Prductlvity difFerent8al.s: which are due to exasrnaiirte% arc n ~ l t I ~ C J U ~ C ~ in O.

as they will h spcifically identified. A drfierentiat~on af eqs. ( I I and (2) yields

where 1 . and I, are lespective xctoral gross investments. I*, and L, ure wctoral changes in labor form, and F, dwribes the margirlal externalit$ eflwt of cxprts on the output of ncsncxprts.

Denoting Gross Domestic Product by Y. and $in= by definition Y ; N + X, it fsllcsws

Using eqs. (3 ) - {J) in cq. 46) yields

&fine total investment 1 (E I , + I , ) and total growth of Idbar t ( FZ L, + Lx). Wmall that eqs. ( 3 ) and (5) imply

Using this result in eq. (7) finally yields

F~llowing arguments similar to those presntd by Bruno (1%8), s u p p s that a linear relationship exists &tween the real marginal prduclivity of l a b r in a given =-tor and average output p r labrer in the wonomy, say

Then, diviJing eq. (9) through by V and denoting Fk=x yields, 3fter some ma~ipulation

The fornorlation in eq. ( 1 1) will k the basis of the empirical work r e p r t d in the next ~ c t i o n . ~ Note that if marginal productiivities are equgliad across =tors ( & S O ) and if there are no inter-wetoral externfalitim (F,=Q%), then q. ( 1 1 ) rduces to the hrnitiar neo-elassieal formulation of the: souras-of-growt h mdel. In the more general cax, the term [b/(B +%a)+ F,] i s likely to k non- zero for le;s developd count ries.

Ifader the prewnt formulatian the parameter should b interpreted as the marginal productivity of capital iri the non-export sector, rather than as the marginal productivity sf capital in the economy as a whole.

&note by TMBK, the total incremerrt to GDP brought about by a mar@nal increas in capital allecated to the export sector; T M P K , can ke referred to as the social marginal productivity of investment in exports, Similarly denote by TMPL, the eflect on GDP of a marginal inereas in export sector labar. Then, one can show

(TMPL, FL)/GL -(TMBK, - Fk)/Gk - 6/(1+ 6) + F,.

Eq. (12) clarifies the interprctatiol; of the term on the extreme right-hand side (which in some of the empirical vyork reported below will be estimated as a fixed pal-ameter): it measures the diflerence h t ~ e e n the marginal contribution to CDP of production factors in the two sectors, relative to the marginal contributions of these factors to export sector's output.

The interpretation of the sources-af-growth equation [eq. (11) ] is then straightforward: the rate of growth of GBP is composed of tile contribution of factor accumulation (i.e., growth of capital and labor) and the gains brought about by shifting factors from a law productivity zctor (nog- exports) to a high real productivity sector (exports).

Econonaetric formulation

Eq. (1 1) was used for a cross-country regression relating "Ihe rate of growth of GDP (in constant prices) to the share of investment in GDP, growth of ppulation (proxy for labor growth) and to the growth of exports (in constant grias) multip1;cd by exports share in GDP.

The study focuses on a group of less developed countries defined by Chenery (1980) as =mi-industrialized. The definition involves both relative and abolute indicators (such as the share of industrial output in GNP and the level of p r capita indu~trial output). Since the indicators do not asesaarily overlap, Chenery (1980) distinguishes between those countries which are only "marginally' ~emi-industriali2ed, and thoye which qualify under 8 stricter definition. Accordingly, the study provides estimates for the mmpk defined by the strict definition (19 countries) as welt as a larger sample f 3 1 ob~rviations) including also narlginal cases.

An issue which neds to k addressed is the length of time to be covered by any singk country obxrvation. Earlier studies used 5-10 year averages, sin= annual dat include substantial random eilects which tend to be eliminated by the procedure of averaging. The existence o? lagged respQnses is another element which hcomes less severe when averages rather thaa

64 (i. Prdrr, On expor!s und ci.onomic grctwrk

annual data are used. 'rhis st,~dy uses averages defined over the decade 1954 1973.

There are a host of econometric problems related to cross-country aggregate analysis of growth (e.g., possibility of simultaneous determination of both dependent and explanatory variables). Some of these complications are discussed by Hagen and Hawrylyshyn (1969) and by Chenery et al. (i970), and will not be repeated here. The constancy of parameters across observations merits, however, some further elaboration. Any cross-country study assumes implicitly that parameters are in some general way similar across countries. In a production function context, different countries are thus assumed to operate with identical production functions. In studies where the production function framework is complicated further by the existence of non-optimal allocation, an additional assumption is invalved, namely, that the degree of misallocation [as indicated by M right-hand side of eq. (12)] is similar across countries.

These problerns need to be borne in mind when parameters are intyrpreted. It is probably better to treat the estimated coeficients as average values which provide a general order of magnitude within the samp?e but are not applicable to any specific country. An attempt will be made in a latter section to allow for a possibility of variation across observations in the externality effect F,. We set out now, however, to estimate eq. ( 1 1) in the form

Y/Y = a . (I/Y) + /~ (UL) + y e ( ( x / x ) . ( X / Y)), (13)

where the parameter y represents the dikrential productivities of ?actors, as explained earlier. I t is expcted that 3, the marginal productivity or capital in the non-export sector, will. be positive but it's magnitude should k ltss than the figures in the range 0.2-0.35 which were reported by ehrlier stildies4 based on an aggregate macro production function. The reason is that in suck studies the estimated parameter is some average (although not necessarily proprly weighted) of marginal productivities in the two sectors, and is thus likely to be higher than a spcific estimate of marginal productivity in the low productivity sector. The hypothesis that margirral productivities in the export zctor are higher and that expr t s generate kneficial ehterflalities suggests that the parameter y should be positive and significantly dlffcrent from zero. The parameter p, related to labor growth. skoald also be significantly more than zero if labor surplus was not the prevalent situation in sample countries during the pr iod covered.

'For example, Humphnes (19761 report% figure\ in the range 1) 2.5 O 99. Pe%mdroglu 4 1972) reprts eritlmrtes In the range 0 .17 0.35.

C;. Fcdar, On exports and cconomie growth

3.1. Empirical results qf the basic formulation

Table 1 reports the results of two specifications of the regression equation: One, reteried to as the conventional neo-classical model, assumes that y =0, and thus present& GDP growth as the result of csipital and labor growth only. The second specification follows eq. (13). Comparison of the two spcifications highlights, therefore, the suprior explanatory pwer of cq. (13). For both samples, the adjusted R* is almost d .tbled when the specifisation allowing for differences nn marginal productivities is used. The results lend strong support to the hypothesis that marginal factor praductivities in the export sector are higher than in the non-export sctor, as the cacfficient of ($/x).(X/Y) is positive and significantly diflerent from wro. The sign and magnitude of the coeficient related to investment are within the range expected. Specifically, when the conventional neo-classical formulation is used, the estimated parameter is within the range observed in earlier studies. When the formulation of eq. (13) is adopted, the parameye- declines sharply.

The parameter associated with labor growth (which, as explained earlier, reflects the relation between marginal labor productivity cnd average output per laborer) is signlfncantly greater than zero.' This may be taken as an

Table 1

Regression results for semi-industrialized LDCs, 19W1973."

Extended sample Limitcd wmple (ii~cluding marginla? cases) (excluding mardnal caws)

Variable Conventional neo-classical (parameter) model Eq. (1 5 )

1: Y ( 0.284 0.178 (4.31 1) (3.542)

Constant -0.P' .. (0.5 54)

Conventional neo-classiwl model

No. of obwrvations 31

"Numkrs in parenthews are t-values.

SWith a 5% one-tail& test.

66 C. Feder, Qn exports and economic growth

indication that surplus labor was not the general case for sample ,- ,antries in the priod under investigation.

The point estimate for y can k used further to provide some inPomation on the social marginal productivity of capital in export (TMPK,). Recalling eq. (1 2), it is noted that TMPK, = (1 -t- F,) . G,, thus

While F, and y were estimated, F , is not known (in a later section of this chapter some direct estimates of F, will be obtained). For a plausible range of F , values one can, however, calculate the corresponding values of the social marginal product of investment in exports, using eq. (14). These are presented in table 2.

Table 2

Possible values of the social marginal product of investment in exports for semi-industrialized LDCs.

P P

T M P K , T M P K , F , (extended sample) (limited sample) - -- 0.10 0.289 0.302 0.20 0.275 0.21'0 0.30 0.264 0.280 8.40 0.255 0.272 0.5?) 0.248 0.265 1.00 0.226 0.243

It should be emphasized that the estimates reported in table 2 are strictly consistent for only one particular (and unknown) value of F,. This follows from the fact that y incorporates Fa, and thus varying the latter while holding y fixed is not strictly legitimate. However, it is evident from these calculation< that the marginal value to the economy from a unit investmen, in exports expansion is substantially higher than that of investment in non- exports. While these numkrs correspond to sample averages, they provide strong support to the view that the siaccess story of export-led economies such as Korea is due ill large part to the enormous shift of resources into the higher prodilctivity export sector.

3.2. Specifyin,? the externality @fleet and empirical results

So far we were not able to decompose the factor productivity differential y into its components. One can idsntify the spcific inter-sect oral externality eRect by adopting a plausible specification for the term F,. Suppse that

exports &fleet the production of non-exports with constant elasticity, i.e.,

N = F(K,, L,, X) = X B #(K,, L,), ( 1 $1

where 8 is a parameter. One can show

Eq. (! 1) can now be rewritten

L 1 L S - -ol.-+p.-+---+o .-.-.- But Y - L I + S X X Y '

Using this result, eq. (17) is rearranged, obtainiilg

x 9 L O)X.x+, - -=.a.-+p.--+ -- Y Y L X Y x'

Note that if it is assumed 6/(1+ S)= 8, the model rcduces to

L I L X - - = a - - + f l - + O . - Y Y L X

which is essentially the equation adopted by Michalopoulos and Jay (14:3), Balassa (1978) and Tyler (198

Results of regressions adopting the specificatiox~ of eq. (18) are reported in table 3.

The modified formulation increases the explanatory power of the model considerably. The results indicate that the inter-sector externality parameter (8) is statistically significant in both samples. The magnitude of the estimated psiamcter is quite substantial. If exports are increased by 10% witho~t withdrawing resources from the non-export sector, the latter grows by approximately 1.3%. The other component of productivity differential ('the parameter 6) can be calculated given the estimirte of 8 and the parameter associated with (%/x) (X/Y). The result is 6 20.75, implying that there is a substantial productivity differential between exports and non-exports in addition to the differen tlal due to externalities.

6The earlier works make a distinction between domestic capital and foreign capital flows. Such a distinction was experimented with in various regressions under the present study, but no (statistically) significant differcna was indicated.

68 Car. P~dgrr, On exports and ac+onomic gpowfh

Tabla 3

Regression results Par semi-iitdastrialiLad LDCs with apcific ~nter-secioial cxlsrnidlit~ec, 19ld 1973.'"

l- - --.__-_ ___-----*----*---- --- -* --- - - ----" - -- - - - - - -- - - Variable: Extended sample 1,imit~d ?camtle (pur::mdter) (including marginal cases) (excluding marglnal caw%)

(a) 0.124 (3.669)

Constant

No. of ubservations 3 1 2 9 ---- ,-, X__s_l_ -_______ --- 7

"Figures in 7arenthcses are t-.values.

The coefficients of investment share and l a ~ o r grotvth arc within the same order of magnitude of the estimates obtained in table 2, and arc statistically significant.'

Using the results of table 3, the social msrginal product of investment in exports (TMPK,) . an be calculated. Recall from eq. (16) F , = 0 - ((1 - x)/x), where x is the shsre of exports in GDP. It follows then,

and eq. (14) can be written

Using the parameters reported irr :,.ble 3, eq. (19) geri~rates the social marginal product of investment in exports for economies with differen; \lalades of x. These are presented in table 4.

The calcl~'lations in table 4 are free of the inconsi~tencies associsted with the numbers presented in table 2. Again, it is Jemoilstrated that, at the margin, investment it1 the export sector has a substantially higher social

'At B 5% one-tailed test.

Social marginal product of investment in exports semi- industrialimd LDCs, 1964-1973. ---------

T M P K , Share cif exports ----- in GDP (x) Extended sample Limited samnle

-- --- 0.1. 0.479 0.512 0.15 0.353 0.41 2 0.20 0.335 0.362 0.25 0.396 0.332 0.30 0.287 0.312 0.3.5 0.273 0.298 0.40 0.263 0.287 0.45 0.255 0.279 0.50 0.248 0.272

marginal productivity than investment in non-exports. The differe9tibil is smaller in economies where a large share of resources is already allocated to the export qector, as in such economies the inter-sector externality effect is smaller.

3.3. Sourccos of growth

Using the . e s u L ~ of table 3 and the mean values of sample variables, the average rate of GDF growth in the period 1964-1( "13 can be decomposed so as to identify the co.ltributions of aggregate in-.estment, labor force growth and resource shifts into exports. This is done in table 5 for both samples. The results indicate that the gain in qductivity due to closer-to-optimal

Table 5

Sources of growth of semi-industrialized J,DCs, 1964-1973.

(2) (3) (1) Parameter Contribution to growth Mean in sample (from table 3) [(1) x (Z)] x lo0

Extended Limited Eh,;nded Limited Extended Limited Variable sample sample sample sample sample sample

------ --- I/ y 0.205 0.222 0.124 0.139 2.54 3.09 ~ J L 0.023 0.020 0.696 0.587 1.60 1.17

I (x/x) (.';,'irj (1.u.22 0.03 0.305 0.302 0.67 0.91 Xlx 0.084 0.105 0.131 0.124 1.10 1.30 Constant 0.006 0.005 0.60 0.53

GDP growth ( p / y ) 0.065 0.040 6.5 7.0

70 (B. I" der , Or6 C S I I O P I S und csconomic growlh

allb,ation associated with export expansicpn contributed more than 2.2 percentage points to the growth of the strictly semi-industrialized counlrics, and close to 1.8 percentage points if the less strict definition of industrialization is used. The contribution of exports can be decomposed into two components: (i) The gain due to beneficial externalities aflccting the non-export sector [which equals 0 . ( I - x) . (x/x)]; (ii) the gain due other elements underlying higher factor productivity in the export sector ((()I( 1 4- 6)) (~1x1 (XI Y)).

The calculations for the extended sample reveal that 0.81 of one percentage pzint is due to inter-sectoral externalities, and 0.96 is due to other effects. Tha corresponding figures for the limited sample are 0.93 and 1.27. Thus, slightly less than half of the gain in growth due to higher factor productivity in exports is due to inter-sectoral externalities.

It is of interest to apply the specification of eq. (18) to a sample of developed economies so as to test for the existence of marginal Pdctor productivity differentials and externality effects. The estimates arc reported in table 6 for both the neo-classical formulation and eq. (18). The results suggest that there is a substantial externality effect, but the extent of marginal factor productivity differentials can not be established, since the coefficient of (x/x).(x/Y) is not statistically significant. Using the point estimate as a basis for calculation yields 6/(1 + 8) = 0.25, which implies (5 = 0.33.

The explanation of the significant difference in the estimated magnitude of the parameter 0 among developed and developing countries requires more detailed analysis, which is beyond the scope of this paper.

Table 6

Regression results for developed economies, 1964- 1973.

Variable (parameter)

Neo-classical Model with externality formulation effects

Constant

R No. of observations

4;. Feder, On exports crnd ec~onomic g r o w ~ l ~ 7 1

4. Coaciudlng remarks

This p a p r pra~lides evidence supporting the view that the success of econoanies which adopt export-oriented policies is due, at least partially, to the Fict that such policies bring the economy closer to an optimal allocation of resources. The estimates show that there are, on average, substantial dicerences in marginal factor productivities between the export and non- export sectors. These differences derive in part from the failure of entrepreneurs to equate marginal factor productivities and in part due to externalities. The latter are generated because the export sector confers positive effects on the productivity in the other sector, but these are no8 reflected in market prices.

The results are such that social marginal productivities are higher in the export-sector, and economies which shift resources into exports will gain more than inward-oriented economies. The empirical findlngs suggest that even when entrepreneurs optimize resource allocation given the prices they face, there are substantial gains to be made due to the externality effects.

The analytical framework developed in this study can be utilized in studies using more tlztailed data such that the extent of productivity differential in specific groups of countries (with different policy orientation) can be assessed. Similarly, !he relation between inter-sector externalities and export composition can be clarified further using the same analytical framework.

Appendix: Sources of data and definitions

A . I . Calculution r$ variables

All data were obtained from World Tables 1980. Variables were calculated from time series for the period 1964-1973 in constant prices. Average rates of growth were obtained by regressing In Z , = a + h e t vvhere Z , is the economic variable under consideration and t is time. The rate of growth, say r, is then calculated as r = eh - 1. Average ratios (investment /GDP, export/GDP) were calculated as simple averages for the decade.

Table A.l presents means and standard deviations of the variables used in the study.

A.2. Composition c$ sumples

Following Chenery (1980), the strict definition of semi-industrialized LDCs applies to the following economies which are included in the 'limited sample': Argentina, Brazil, Chile, Taiwan (China), Colombia, Costa Rica, Greece, Hong Kong, Israel, Korea, Malaysia, Mexico, Portugal, Singapore, South Africa, Spain, Turkey, Uruguay and Yugoslavia. In additior~, Ireland and Venezuela fall under the strict definition but were excluded from the sample;

<'h;~rauter~s~ifi of the samples of semi-industria1izeci 1.11Cs. 1 9 M 1973.

Mean (standard dcv~ation) "

Extended Limited sample sample:

GI9I' growth O.CM5 (0.022)

Investment (il9l" 0.205 (0.050)

I'opiilation growth 0.023 (O.(W)9)

Exports growth 0.084 (0.070)

Export share 0.235 (0.225)

(Iiuport grobth) x 0.022 export share (0.032)

the former is defined by the World Bank as a developed country and the latter is a tnajor oil exporter. Countries which are considered as marginally semi-industrialized were added to the limited sample creating the 'extended sample'. These countries are: Dominican Republic, Ecuador. Egypt, Guatemala. India, Ivory Coast, Kenya, Morocco, Peru, Philippines, Syria. Thailand and Tunisia. In addition, Iran, Iraq and Algeria are defined as marginal cases but were excluded from the sample, being major oil exporters.

The sample of developed countries consists of Australia, Austria, Belgium, Canada. Denmark, Finland, France, Germany, Ireland, Italy, Japan, The Netherlands, Nc~rway. Sweden, Switzerland, U.K., and U.S.A.

Hdld\\a. Bela. 1977. Pol~cy reform In developr'lg countries (Perpamon Press. New York). Bnldwt. Nela. 1978. Expoit\ and economrc glowttr: Eurther evidence. Journal of Development

1-coriom~c\ 5, no. 2. June, 181 189. Bh'igwati. Jagdlsh N. and T N Sr~n~va\an, 1978. Trade policy and development. In. Rudigcr

13c~rnk~r\cIl ,inJ J,rioh 4 1 rrnkel. r J \ . Intern,itloniil economic pol~cq. Theor) .tnd rk~dence. ch 1 (John\ Mopkins t;n~vcrctlty Press, Baltimore. %fD) 1-38.

Bruno. %lich6icl, 1968. E\t~mdtion of factor contribution to growth under ~tructural dicequlllhrlum. Intcrnatrondl F.conomtc Review. Feb.

Chener, lioll~s N., 1980, The wm~-~ndufi!rlal countries, Draft, July (World fianli. Waqhlnpton. I X ' )

Chenery. biollls R , Hazel Flk~npron and Chr~%topher S I ~ \ . 1970. 1 uniform andl>\~r of dctelnpment pattern\. Iconcjmic I9ekefopment report no 148. J u l I l larc~rd C'nlters~t). Cdtn hrtcipe. MA]

)4;9gen, 81. Everett and 011 Hawrylyshyn, l%9, Anal!isis of world Incrlmca? attd grctwth, 1955 IYb5, Ecclnr~rnlc Ikvelopment and Cultural Change 18, no. 1 , Part El, Ocl,, 1 (30.

Weller, Peter S, and R~chard C. Porter, I9'98, Exports and growth: An empirtcial re-lnve8tlg&r!on, Journal of P)i.velopmcnt I:conomics 5, nil. 2, June, 191 193.

I-tr~mphnes. Jarle. 19"IW, Causes cf growth, Economic hvelopment and Cultural Change 24, no. 2, Jan.. 339 353.

Kecslng. Donald B., 1967, Outward-look~ng policies and etwc9nomic develapment, Econom~c Journal 77, nu. 3CM, June, 3Q3 326.

Kes~ng , EPonald B., 1999, Trade pulicy for developing cotnntnes, World Bank Staff working pswr no. 393. Aug.

Krab14, I r~ lng B , 1940, Trucie us a handnialden of grouth S I I T ~ I : ~ P I ~ I ~ ~ k t ~ ~ e e l i !he n~neteenrh and tv\lcntteth cetlturies. Economic Journal SO, no. 120, Dec., 8%) 872.

Krucrger, Anne 0.. 198Q. Trade plicy as an ~nput to development, Amer~can Econ~~nitc Reblew 70, no. 2. May, 288-292.

Michael!. Michael, 1977, Exprts and growth: An emprrrcal tnvesttgatlon. Journal isf &velopment Economics 4, no. I , March, 49 53.

Mickalopulos, Constantine and Keith Jay, 1973, Grouth of exports and income In the developing world: A neoclassical view, A.I.D. d~wussion pawr no. 28 (Washington. W).

Pesmazoglu, John, 1972, Growth. Investment and savtng rarlos: Some long and med~urn term sssoaatronr by groups of countries, Bulletln of Oxft~rd linlbersity Instltic!~ of Econorn~cs and Sta t tu t i~~ 34. no. 4, Nov.. 3U9 328.

Robinson. Sherman. 1971, Sources of growth in less cievelopd countries, Quartcrl! Journal of Economics 89. no. 3, Aug., 3 9 1 4 8 .

Tyler, W~lltam. 1981, Growth nnd export expansion in debeloping cc\untnel. Some emp~rical ebtdence, Journal of &velopment Econornlcs 9. no. 3, Aug . 111 130

VC'c~rld Bank, 1980, World tables (Washington, TIC).

C;t%rsh~t~t t:e~4rar, RtctP&~r(A qu?sf, df~ci Knud ROSS, 'TProle"ting I jcbt S C ~ P V ~ C I G P ~ G'apdcttv of L ) c ~ v c * ! z ~ ~ I I " I ~ C'ctci~tfrit~~," f o t ~ r l t , ~ l c z t I / ) t r l t l r i t$ / l ~ ~ r g d

a ) l l i / l l \ / f f l t l t t t h \ l f l k ! ~ * l ~

Hta-h,irci F I . C;crldmar~ dnci L,yn S q u ~ r ~ , "7"et-hrttcal Change, I+abtlr Use, drsd ~I ' ICOIPIC~ I l ~ s t r l h u t i o n IFP the M u d & Irrtgatlon IPrcpltr.~f," I t c ~ ~ t r ~ t ~ ~ l i

)c~*c"lr~/~l)la'!d 111111 Csl l ! f llrrtj CT1t l~f t$t '

M l c h ~ t 4 lE:~ngvr, "'X'rrlde .'~nd the Structure* of Americdr~ lrsdustrv," ,\ rlllrl!s c r f f i r (? t6\11t(*r~('iif1 ~ J ~ ~ ~ T F I I Y ot / ' O ! I I I C t$/ nrtci r l s l Sr tc*trc I *

l3dvtd R4.C;. N t ~ w b ~ r y %and Josc~ph E. S t i g l i t ~ , ""Optimal Cornrrsodity 5ttta.k-pill ng Rulcas," C?p)(rrrI C.r-r~rltlt?llc I\t)tf9rs

f" N S F I I P I V ~ S ~ ~ P , "C;~anerdl E q u l l ~ b r ~ u m 'Theory, Prolect Evaluatton, a ~ s d Eiconomlc F 'Pt~vt~lopn~t~nt , ' I rJit-Tlttv~(trq u ~ I ~ I P ~j?t*rltvtli 6' t ~ / I t i~ t~~vn!f i I~c*i~r*loj~fifiv""t

T'r~lmarsurl J i r n c ~ r ~ c ~ / , "'ripe Vdlua. of Squa t te r Dwe111ngs lss Developlr~g C cru n tries," f c o r r o r r r r c I )c*ilt*lt~;)rrrr~tII mtrd C"liilfurtrl Clrurr$t* Boris PItaskovic .-d Mc~r jan I3clienc, "'Regional Developmabn t rn a Soc~,llrst, Dclveloping, . ~ n d Multrnaticr~l~11 C'ountm. The Case o f Yugo- \ l a v ~ a , " /~rrft.rrlritrc~rrnl Itr*grorral P;I rc9?tr r* Rt*zsrc3zc~

Mt,-l\c) N l s h l m i ~ u a n d J i thn M 13age, jr , "Tcjtmi I-actor i'"n~ducllv;lty C;rowth, ~Tcbchnc~lc)gtcdl P r c ~ g r ~ s s and Techngcal E:t%lcicancy Chcangrg Drmvnstons of I"soiuct~vlty C'hdrnge ~n Yuga?sla~!a, IQh5-78,'' T!!t"

j M h-~r.rgt*r, " 'The% Pt~liflcdl I<conomv of Adm1nistere.d Prnta~ct~crn"' (v. 8th k 1 Kcqlfh Fi&~. , l l c~nci I3ouglas R Sdvlaon 1, T l r t * /*irtttgrrtnrj I I otaomrt Rt~~lrc , i i* , 'ZIIC~ "The Industrv-( 'ountry Prsc~da~ncc~ o t lLt-*s% than i-alr "*'d?uee C ases t n L' S Import I'rsjctt*," Qiiizrf l ~ r / 4 Rt*r*j:*cc' c l t f r r)ttcPpPlir 3 rlrtai R ~ s l ~ t r s s

ktlmal I l ~ r v ~ s , J a ~ m t " C ~ C " Mtbli~, a n d S l ~ r r m a n K o b t n \ t ~ n , "A Gc~r-reral I 'qudibr lum A n d l v s ~ s of I ;orc l~n E\c?lcongc' 5ho~t~agc . s tn a I3r-vcloplng ij conc?mv." T l r t , I ( ljrlflnrrt fc~trrrlul

K\iu 5 1 k ! . e t a , "A Model of I r ~ t r ~ ~ u r k a n f-mplovment L crcdfltlr'f A n i1ppllc~fic111 Rogcjf~, Cc~lornhr,~." \iclarrirrrl c j t I l r b $ r ~ t l i ' r rltttlnrrt %

I W Knight ,rnd R H SaC~ot, ""Irom Migrants lo Proletdrians f A r n p l ( ~ v m c ~ k,,x)gt3rlVncc., M o k ~ l l tv , drtd bLt'r~at"% 111 T1~n/dnlC~," 01 ford R ~ r l l t . l r r r r l f T (cttrotlrrt s t r r r t B \frzfr*frt.

M I,cturse ftclx, ""incomcl Fllstributlon if:: Pi~st-lQh-5- Hr,l/nl N v w Ktasu Its," Joltrtl( j l ( f / 1 t j ~ t o ) t ? l f / l l h f o r y

'"B"ME3 IWRLD BANK

I it.sdq"i8rf ''5

1818 H Skaeer, N.W,, Washingkcan, D.C. 20433, U.S.A.

Teieykone. (202) 477-1 234 Telex. W U l 64145 WCIRLEIRANK

RCA 248.123 WC)RL.DBK Cssble address: INTBAFKAI3

WASI4ING'I"Q3NI)C

F st rrtptsrcrtr C)@c t s

4.6, avenue d'lrna 751 16 Parts, f'rancr

Telephone: (03) 214-5W3 TeIex 78 1-26838

The far!! range of Wrld b n k publ~siatton~, &>th free and for wle, 1s dexrrbed In the Cwlahty rrf Puhlttatrc9,:s; the contlnurng T W A K ~ prc~gram 1s outlined la i-liksfrstta of Clrrre~l S f u d r ~ \ The mcmt rmealt edition of eactt 1s avarlabie wrthout charges frr~m

PU WLICWTKrNS D1STWi WU'ffON UNIT THE: MPWLI3 I A N k

181s H STREET F6 W

WA%t.lrlNC;TON I1 C ;t13a33

v s a