POLICY RESEARCH WORKING PAPER
Product Quality, Productive Efficiency, and International Technology Diffusion Evidence from Plant-Level Panel Data Aart Kraay Isidro Soloaga James Tybout
‘The World Bank
Development Research Group ‘Macroeconomics and Growth and Trade January 2002 Wps 2154 2759 oxocc, fs past exposure to orelgn snowteage through expats imparted
Trang 2Pouicy Reseancis Wonkise Pavan 2759
‘Summary findings
‘Wat mechaisns mow fegueny ans foreign technologies to developing curty fers? Do these foreign technolo affect both prodctiveefciency and produce qui the recipient fs? Under what ‘Greumstances do fms pore aceite tha iether, ascent foreign knowledge? “To ailres these questions, Kraay, Solow, and "Tyhou develop anew methodology and apply the framework 0 plant eel panel dt from Calbia Mexico, and Moroso Thee ests point sever sic mesg, as, by imposing enough ractare onthe production
function and the demand syste, itis poser measure rodut quality and marginal ox athe plac eel and toelat the evolton of these vase to it acy histories Doing so, the aor fd eons feel perstence in both quale and marginal costs, Ba in moc indy o county panels ha they stupas iceenatonal ates hep ide in prediing cuerent peformance ence pst relations on gui and ‘marginal cos are contol fr That ix aces donot ‘ypily Granger ete performance neering, in the minority of ses where signican aoeiation,
hoạn, intersational aces appear to move costs abd Produc quality in he same drecion So the ne fect, ‘of profits in hee cn nor immediately apparent Second, several tase pater emerge wit spec 10 che đosemimmng of international actives, Mont fundamental, acinar highly prsen, even ser ‘unobserved heterogeneity icone for Thar sugges ‘hat ir incur unk theo conte when they nian or cease acre 0 remporay poly oF Imseroeconomic shocks may hav longum effects onthe tera of actives observed in prt counry or Ide Also cies ed ogo toes shat sae tha telat frs performance to one iterations activity bd igore de ors may geneate mieading ‘oncasion Bur dhe bundling of aces seems 20 ‘aay refer unobserved plant characteris sch ‘managerial philosophy, comtacs product niche, and Iocan One these are contol fo, hte i cridenc that engaging in on nerationlaciity Increases the prbabiiy hat frm wil engage in obers inthe fore
“This paper—a june product ofthe Macroeconomics and Growth Team and the Trade Tea, Development Resear Groupispart of large llr he group to stad pater ofnterasonal technology diffusion The sty wasfonded bythe Bank’ Research Support Budget ande the eesearch project “Micro Foundations of ternational Technolgy Dison Cpiesof his paper ste aval fe fom the Word Bank, 1818 H Steet NW, Washington DC20833- Please contact Rina Bonfield, room MC3-384, telephone 202-473-1248, fax 202-522-3518, email address shoafeldawordhankorg Poly Research WotkingPapesateaso posed onthe Web at hepieconrldbank.og The ‘suthos may becomacted stskaay@wordaak or, Soap @otma com, orjst32@p.ed,Jastary 2002 (pages)
‘See her
Tie Rly Roach Wong Pap Sa an te dg ow pan ge eae of aa Lrvopan a coy hen he hn on hn be eg Tei eprom a so ee choo ngage te prt el ped Ta
‘srt tnt tr They Sa mel nt ie to or Bn 8 at Dc
Trang 3Product Quality, Produetive Efficiency, and International Technology Diffusion: Evidence from Plant-Level Panel Data ‘Aart Kraay ‘The World Bank Isidro Soloaga “The World Bank James Tybout
Pennsylvania Sate University and NBER
‘This paper i part ofthe World Bank-funded research project “Micro Foundations of| International Technology Difusion.” We thank Welfgang Keller, Jim Levinson, Mare Melig and Mark Roberts for useful discussions and absolve them of blame for
Trang 51 Overview
‘A casual tourist can confirm thal ecologies olen make her way rom the eveloped word to theless developed counties (LDCs) However, despite the critic importance of technology diffsion for economic developmen, the evidence on many
aspects ofthis process remains sketchy: What mechanisms most fequently transmit foreign technologies to LDC firms? Do these foreign technologies affect both productive ficiency and product quality inthe recipient firms? Under what circumstances do fms pursue activites tha ghe them acess to foreign knowledge? This paper develops anew methodology for adresing these issues and applies the framework to plan-level panel
data from Colombia, Mexico and Morocco
A ‘The existing Wierature
(Our init understanding of ntemational knowledge diffision doesnot derive ‘om neples ofthe topic nthe empirical economics iteratur alone at east three
<Sfferent methodological approaches have been deployed Fit a the very micro level a umber of analysts have used casestudies and qualitative survey to generate
descriptions of leaming processes at individual firms ex, Hobday, 1995; all, 1987; Katz, 1987; Pack, 1987; Rhee, Rost-Larson and Purl, 1984) As Pak (1998) observes, this literature provides invaluable details concerning ims" effort to absor technology Đo ha ote to say quantitatively about he results of these efforts in terms of odustive efficiency or product quality There ar some exceptions (¢g, Pack, 1987), but they are too few and based on such smal samples that they provide ile bass for
Trang 6Atte other extreme, studies based on aggregated data have coelted coss- county patterns of productivity growth or productivity levels wih various proxies for counties exposure to forcign knowledge andlor thế ah to absorb it These proxies include apt goods imports (¢, de Long and Summers, 1991; Keller, 2000), ade ‘with countries possessing large RAD stock (Co and Helpman 1995; Keller, 1998;
Keller 2000), foreign dirs investment inflows (Blomstrom, Lipsey and Zan, 199), and domestic patent stock (Eaton and Kortum, 1996) Unlike case stuties and <escipive surveys, cross-country regressions document broad patterns of association, ‘and in tha sense they provide a basis for generalization However, most are subject
‘avity of econometie eritcisms, including ageregation bias, omitted variable bias, measurement eor bias, and sultant bias
Finally, plant of frm-level econometic stdies coelate proxies for firms! exposure to forign knowledge with thst productivity levels or growth aes To ita fw examples, Aitken and Harrison (1999), Haddad and arson (1993), and Djankov and Hoekman 2000) study foreign diet investment and Chen and Tang (1987), Av and Hwang (1995), Clerides ea (1998), Kraay (1998), and Bigsten eta (1999) study
exports? By sacrificing the nuance and dtl provided by casestudies, these miro economic studies gain the ability to weatlrge numbers of producer and make steal inferences They also do beter thn the macro stuiies i ems of denying
the specific conelates of productivity and avoiding aggregation bas
Trang 7
Nonetheless, the plan1evel eeonomeie dudies hay signiieantshodeortings too, One common problem isa filure to disentangle causality For example,
contemporaneous coreation between exports and efficiency tells us nothing about what ‘caused what Bven studies that use lagged exports to predict current efficiency may miss the knowledge transmission mechanism, as Westphal (2001) has emphasized A second ‘problem is that productivity is almost always poorly measured.” Manufactured products re quite heterogeneous, even within narrowly defined industeies, so there is no single
measure of output that can be compared across firms Real revenue is typically used as 8 stand-in for physial product, but this variable responds to product-specific pice
adjustments as wel as fluctuations in physical volume Productivity measures ‘consequently confound productive efficiency and market power Further, when technology ffusion leads to product innovation rather than process innovation, these productivity measures may miss the effet entirely Finally, lke the cross-country regressions, studies in this literature usually Focus ona single conduit for international technology difusion and ignore the others, opening the door to omitted variable bas Đ — Theeontibutonof
paper
‘This pape is an econometric study based on plant-evel dat, oi fs squarely {nthe third category mentioned above, But we atempt to improve on existing
methodologies in to respects Firs, we abandon the standard approach to measuring productivity in favor ofan altemative approach that eats the observed plant-fvel data > ew sus in he ter teats do beter bu Pack (1987 is an exception He ses deta information on machines aed workers aa ample of plant caleulte high quality preety
Trang 8on revenues, costs and market shares as reflecting equilibrium ina differentiated product market By using @ new normalization and imposing suficient demand-side structure, we avoid the problem of distinguishing real evenue from physical product We are alo able to separately measure process innovations, which are manifest in marginal cost
reductions, and product innovations, which are manifest in heightened demand fora product ata given vector goods prices Accordingly, we can study the joint evolution of these provesses and investigate whether improvements in one dimension are
complemented or offset by changes in the other
Second, we teat multiple channels for intemational technology diffusion in a single integrated famework This would not be important if he various activities that transmit technology were unrelated to one another But because of complementartes and inivisibilites, they tend to come in bundles andor in predictable sequences" Hence econometric models that treat any one of them asthe unique source of foreign technology run considerable risk of misatribution,
‘Table 1 below lists the main knowledge-transmiting activites identified by the case study literature, Its not hard to identify reasons why firms’ decisions regarding ‘these activites will be related to one another Exporters are relatively likey to use
imported capital and intermediate goods because they ae granted preferential access to forcign exchange, or because the produet characteristics needed for exporting are best manufactured with these goods Similar input and capital god requirements may
Trang 9seeompany lieensing agesmeni Fima ih FDI are relatively likely to use imported intermedtes because the parent company can nteralize some costs by doing so Multinationals sometimes locate plans sbrod to exploit low wages wile protecting thee intangible assets ike poprctary knowledge and product reputation, so FDI and
exporting can also be complementary activites Table Alternative ways to acquire foreign technologies Foreign direct investment in domestic enterprises; 4 joint ventures; outsourcing: ‘+ fcensing arrangements;
‘+ importation of intermediate and capital goods, + leaming from exporting to knowledgeable buyers;
‘leaning from final goods imported, and reaetions structures as these goods enter the country to changes in domestic market HH Methodology
A The Conceptual Framework
Trang 10Performance determinants:
Let us begin with a representation of performance determinants From the
perspective of managers, the activities in Table 1 generate revenues by improving a mm uit production costs (c,) rte appeal of ts product (a), which we combine to fom the performance vector @,+(ay,¢,) forthe j*finm in period Then, presuming
thatthe elements of ey, evolve according to auto-sepressive processes, condoned on «exogenous frm characterises (hereafter) andthe firm's history of activity bundles (hereafter B, ` @® where ,= (0,94, 6s), B,
lBycb,B, si dh Ey isa serially
uncomelied vector of unobserved innovations in theo, proces, nde clus ved of tummies 6, indicaes wich stv bunle the fem is panna tine iany*
Act đươnnlomnn
(fours, he vies thaseves are endogenos, We envision ams weighing four kindof fects on thei prof eens when choosing which combination ope, Fina desi by equation (1), stv in une are reizationson te
pevtmane tection in so ding thy may as ae ther atviy
choices and performance of competing firms Third, given the ew, realization, activity
5 hats the A cement of by tke vale of oe in pevod ithe frm i enaged inthe È*
Trang 11bundles can affet net operating profits by changing demand conditions—e g., providing access to foreign markets—or by affecting the share of operating profits retained by the firm's majority owners Finally, the initiation of activities generally involves star-up or adjustment costs
‘The speifis of these effects on profits depend upon the activity in question For example, joint ventures, subcontracting and FDI may transmit knowledge andr improve ‘firm's access to inputs, thereby affecting the evolution ofits performance vector, ,
These activities may also affect its operating profits by ereating new markets for is products, by branding its products, by creating profit-sharing obligations, andor by <iluing comporate control Finally, joint ventures, subcontracting and FDI involve up- front research costs and legal fees when they are initiated
Similar observations apply to the other activities in tale 1 Firms that import Jnermediate or capital goods improve their performance trajectories by using higher «quality inputs and by extracting knowledge from these foreign goods But they also incur higher material or capital ests and, prior fo importing, they must research foreign suppliers and leam about customs procedures Fis that employ high quality workers ‘ypically improve ther processes and their products, but they also incur higher labor costs, and they bear the sunk costs of attracting and screening job applicants for these
Trang 12‘When managers understand these linkages and corerly anicipate he behzvior of their rivals, the activity choices ofthe firm in period canbe represented as
determined by the decision rule:
by = BZ X My BT y) @
Hore the arguments of g,() include everything that helps firms predict the future pay offs fom each possible bundle: cutent and past exchange rates and demand levels, Z, the set of previous realizations on « forall industry participants, (O35 p45 ye) Previous activity choices forall industry participants, By
(Bye Byy"s Bry) ste exogenous characteristics ofall industry
pedidpenl, X, = (Xu so iXie,); na seofbelcB, [,,sboạl the dessin
rules that wil be used by all ofthe other firms (When the industry sin equilib, ‘these beliefs must be consistent with observed behavior) Finally, while the information
setis common toll firms, g,() depends upon j because the / firm's own characerstios and history affect its pay-offs asymmetrically from those of ll other firms
Inference
(Our basic objective isto quantify the relationships described by equations (I) and (2), We shall view significant associations between B,., and a, in equation (1) as ‘evidence thatthe international activities Granger-cause performance Similarly, when
Trang 13<etermine whether particular combinations of bundles are elaively potent performance ‘determinants, Also, given that we shall examine the join evoltion of productive
ficiency and product quality, we wil be able to make inferences about the nate of the performance effects induced by international contacts
‘Given sufficient variation in the data, this approach inference should pick up ‘most instances where knowledge soquired through observable activites enhances future
product quality o productive efficiency However, thee ar some types of linkages ‘between activities end performance that it wll fil to detet For example, suppose a foreign corporation subcontracts wih particular plant to bocome a supplier fr one its products andi transmis the necessarily technical information to that plant tmay be
yeas before the plant actully begins to export the produc to he buyer Ce Kim, 1997) so when the associated exports show up inthe dat, the plants performance tjectory
will ave already responded to the new knowledge and no association wil be detected, In fact, this scenario would generate the misleading econometric impression thatthe
coulsourcing stvity responded toa productivity shock rather thn vce versa
qu om (1) wil lo miss echnology diffision that does nat cura the fmm level For example, if ims acquire imported intermediate or capital nputs through an intermediary rather than by purchasing directly frm foreign supplies, the associated improvement in performance wil ot be tibued to foreign sources Silay, if finns that eam by engaging in international busines serve as valuable examples for others, the knowledge spillovers they generate wil not be attibued to foreign sources For all of these reasons the results ofthe exercise hat fllows shouldbe viewed as suggestive
Trang 14Đ AnEmpirealModel
‘To render equations (1) and (2) empirically wseful, we must surmount several obstacles Our first problem isto measure the performance vector, @ Somehow we must ‘measure unit production costs (¢,) and produc attractiveness (ay) concepts using plant-
level data on revenues, intermediate input costs, market shares, labor costs, and crude capital stock proxies
‘To render unit production costs observable we use a novel normalization ‘Specifically, we define one uit ofthe * plant's product o be whatever that plant can produce witha dollar's worth of intermediate inputs, Since products are differentiated ‘this doesnot imply that a plant using a elaively large amount of intermediate inputs is ‘producing a relatively valuable outpat However, if firms were to differ in the efficiency
with which they convert intermediate goods into final output, this assumption would have the undesirable implication that efficiency guns reduce output when the physical volume of final production doesnt change * Thus our normalization is accompanied by two strong assumptions: fins exhibit constant returns homathetic technologies, and these
technologies differ across producers only because of diferences in primary factor
‘efficiency Tat is, some firms use labor and capital more effectively than others, and this {is the only reason marginal cost schedules differ across firms
‘With these assumptions we can calculate unit (variable) production costs and
© Varian in factor prices ars plats wl regional variation nator psig regional dames ei ‘Sopped the forthe es reprtd Here,
for inference, We cold for ound spies effect so we
Trang 15\atiable costs (labor, intermediates and energy, /, be intermediate input costs ad 8, te revenues Our assumptions also imply tha unit variable production cost correspond to marginal cost
‘Our approach to measuring product ppel or quality is more involved Iti based onthe notion that, given the vector of prices forall avilable products in an industry (Gocluding a composite imported variety), large market shares imply high quality OF
course, market shares reflect more than product characteristics, so one shovld think of “quality” sa broad measure of produt appeal that responds to reputational effets and
advensing as well as physical characteristics ofthe products
To impute this quality notion fom prices and market shares we needa demand system and a markt oxulibrium concept For these we us La and Tybout’s (2000) aptation of Bey (1994) representation of a diferente product matket, which in ‘tums based on MeFadden's (1974) nested logit demand system ad the generalizations <eveloped by Berry, Levinsohn and Pakes (1995), The following paragraphs paraphrase {Luand Tybout’s (2000) deployment of Berry's (194) model
‘The Demand System
We begin by assigning each producer in the industry of interest to one of G aengraphi region (nests) Producers in ll regions compte wih one anther and with « composite imported good, but consumers view products within a region as closer srbslles thin prods coming from distin regions The price a he composite ‘imported good is exogenously determined by the weal exchange rae There ae N
domestic establishments, indexed by je (1 V}, each supplying its own unique variety
Trang 16
‘varieties Finally, © she set of produc varieties inched in product's nest isting prods
‘Domestic consumers have heterogeneous tases, indexed by the real number 46 (.L,] Bach period, each consumer inthe market chooses a single unit ofthe variety that yields him o ber the largest net indirect tity, where variety yields consumer € net tly: #6, Mo ® Hang,
1 Py forthe N domestic vaveties and iy = Ey ~Y-r, fo the imported
varity, where &, exes the “gait” of good and p and arte prices ofthe donssde gosds andthe domestic cuency price ofthe imported composite good, renpectvly
‘The lst wo tes on the right hand side of 2) re unsbserved eno components that capture individual taste difeences The ist component vres across nets but
sot within them, while v, exhibits withinnestvaraon By assumption, bth
[¢ +v]andy are distributed type-I extreme value across consumers, with variances
(6) /3 and (13a) 13 respectively The inet tility funtion parameters
‘are identified onl up toa scalar multiple so we impose 4, = 1.” Also, we define “Anderson, Palin un Tis (192) show that on can hak ofthese cor components a ‘efleingbeterogeneous ste over unaburved produc chances From at pnp,
‘Ackerbrg a Ryman (2001) sage tat nj be ade econ ofthe numberof rods, ‘sce the adios of mre pout a markets helt crowd product sac and eflecivey reduce {he nperoa nts cron products We experimented wit ks geeralation sound 20 cs
Trang 17
L4 1/4 (8< <)) to index the ageeof subsiuahilty aneng,venus witha, the nes!
Integrating over domestic consumes yes the standard ested logit expression forthe demand forthe domestic variety a fraction of total domeste demand for vatites in the product's nest exo, ~fy)1-0) Dovel, Hyer) @ Similaly,toal đemand or grop g vrielietas a share of total domestic consumption is G8) " )` -Hy)M-0], B= Sore %
and the demand for the imported varity as a share of total domestic consumption, the import penetration ato—is hư gò Hence demand forthe * variety asa action of total units sod is 5, — Given our normalization rule, we can measure the market shares that appear in Dh ẨM ands, 4), (6) and (9) 98 5, M Sayer where My Dh
Trang 18years? Then, using these equations and the defini of Hits possibleto solve forthe ‘quality of domestic ood /by using the expressions for mean uty
Finely, we obtain our quality/ppeal mesure, a, by expressing, relative othe
quality of imports Specifically, without loss of generality we st the mean utility from Imports to zero (i, = 0) and we measure the quality of domestic good j relative to the quality of imports as:
ay Ey ©
@, "`
[Note that equation (6) doesnot explain how a, is determined rather provides away to solve for an unobserved matix in tems of observed matrices and parameters that canbe estimated,
Estimation
‘OF course, equation (6) cannot be used to impute quality unless ar, yand o-are known We identify these parameters by substituting the right-hand sie of equation (6)
ito a linearized version of equation (I) wherever a, appears:
More presi, we calle, sing a constant el exchange rte 1 convert el dlr ipo o ` finemediate ponds ued to produce tem Our aproch o messing imports als implies a foreign produces do ot aut idl rc in eponse to exchange ate Nacaton—th 6, ‘we sre complete ps-rovp This impletion consent vi ur expressions Ist ses and pies, wch re sed nthe premise tha be price ofthe use good doesnot fer qm
‘eond eds he pts of te domestllyprodieed vis, Whe he epic Feature suggerts are dlr pices of imported gods ariel to speed soe o exchange ale ‘uetanons (Knee and Goldberg 1999), or essumpton es tot seem oa fom aly
Trang 19-HÊ nh t mc tôn cho Hy a DA +05, a) Bhs 2 oN, te Qt
‘Then we estimate these equations jointly with the equilibrium price relationship that ‘obiains when firms compete Betrand-Nash inthe prodvet market (Bery, 1994: ©
woolen ee | ch lu o
“Tavis we esinat th demand parame ath sane tine hat we estate pecans desing theft ofintematna seve on ech dimension of
performance (4 and ¢*)
iit win pi a tal lth Nad
(2) First, if reflects measurement eror in costs or intermediate inputs it wil be correlated with the right-hand side variables in equation (7) Unlike Berry (1994), we assume this problem away by positing thatthe noise in equation (7) comes from exclusively from measurement error in revenues (and thus in prices)
‘Thisconion presumes that te (common know) petformance Year (yy, Oy ‘is reaiae tte bean of ech prod and that fae tse te indus do at depend pon cet rice and tay cies, given, mo, Alb, mined sates inthe prods ‘nuke competition ‘eltinbetven prac mtt competion and indy dams) ae dislowed (Se ieae a Pale (195) fra formal discussion of the
Trang 20Second, ifthe disturbance terms e7 and Z7 are anidly corelated they vi not be orthogonal to lagged endogenous variables (4, Iyar€y-q) ad spurious corelation
‘ates may result Ther rc two standard ways ode wth his problem One to choose a sufficient lon la length (0) that al persistence inthe endogenous variables
is absorbed bythe explanatory variables, leaving 6 and serially uncomlated This solution is simple and appealing, but in short panel ke ours it means sacrificing most of ‘the ime series information in the data Since time series variation is key for Granger
causality tests, we choos the oer standard approach Thats, we assume thatthe eror terms are characterized by a standard erro component specification and we eorect for the associated inital conditions problem (Heckman, 1981)
Specifically, we write the disturbances as €% = pf +0} and ef = uj +05,
where-varoh)=22,,varlut)=o3,, cov ) 20 We, eov(o3,0}
cow 5) = om a cov
‘Wem and VE = ae Thea wwe ute Wooldidge’s (2000) conditional likelihood function to deal with the initial
sondidons problem Tats, we write the joint density for Tealizations a the veto 0 andthe unobserved effets, 1, = (15), conditioned on and x,
SO ty)”
LO Opty 5p hyp) Igy ip kph si SO Oe k shy )80 l8,
‘Then we express 1,88 a inear projection on and the temporal mean of x, plus a residual plant effect:
y= Bo ay + BE, +H ®
Trang 21nly, we subsite his expression for 1, into th density fron shove conioned on ands, and weintegrate out the unobserved plat ets = a) Since sis onogonal oy by consration spot imines te inal condions problem andthe compound disturbance vec, f.05)+ has andar cư
‘components properties." We use a fll informetion maximum likelihood estimator forthe system (1a), (1b), (6), (7), and (8) presuming that al disturbances are normally
distributed
‘Our final methodological task is to develop a version of equation (2) that ean be ‘estimated Stuctural estimation ofthe deep parameters is out of the question, given the ‘complexity ofthe optimization problem and the numberof parameters involved Instead,
we assume thatthe probability of choosing the # activity bundle canbe writen as 2 reduced-form linear expression inthe observable arguments of g,() Also we drop all Jags of more than one year and we summarize the performance of competing firms with & cross-firm average of o, excluding the /* frm, hereafter 7
bw =ố, By, 4 Oy, HN l2“ @)
‘Noe tht we ae aru g's dependent fp ugh ony not be,
Trang 22(Unlike in equation (2), firms’ choiees of activity bundles are not deterministic here ‘because we do not have access to the entire information tha they base their decision spon.)
‘As with equations (Ia) and (1b), we adopt an error components specification for
the dituebances, c=} +0, and we addres the associated inital conditions problem ‘sing Wooldige 2000)s technique, Thats, we expres ach clement of the vector
as linear function ofthe vector of initial states by» plus noise
Inprincple there are some efficiency gains tobe reaped by estimating the sytem ‘of equations 2" jointly with those in the system (Ia), (1b) and (7) above, However, so
long asthe residual plant effects in equation (2) are no corelated with 1” in equations (19) and (1b, this
not necessary for consistency We therefore opt keep the model ‘manageable and estimate the activity determinant equation separately ‘Empirical Results ‘To implement our estimator, we shall exploit annual industial survey dat fom ‘woe semi-industratized counties: Colombia, Mexico and Morocco (See Roberts and
“Tybout, 1996, for detailed descriptions of the dat Ideally we would like to sty al of the activites dened in able 1, but unfortunately, each data set provides information on only a subset, The Colombian data reveal wheter firms are exporting and whether ‘hey are importing intermediate gods; the Moroccan data identify expos and firms ‘with foreign owners (FDI), andthe Mexican data identify exporters, importers of
Trang 23spon the ere set of activities, tween them we observe a uty large colleton of «conduits for international technology transfer
For several reasons, we shall fews ou mpivical analysis on manufactured chemicals Fest, to teal ofthe establishments in our datasets would be 2
overwhelming ask Second, among the sectors with suficlent observation o support
inference, the chemicals industies are relatively prone to engage in international
sctiviies, Thin, these indies also rely cativey heavily on scientists, technicians and cngineers, so when technology diffusion takes place, we ar likey to find it among them Fourth, in most chemicals industries, imported final gods are sufficiently important to play the role ofan ouside goed in our demand system Finally, the chemicals industies are wel represented in each of the countries, Ths cos-country comparisons allow sto examine wheter particular types of prodtion are prone to particular paters of
technology absorption,
‘We sal lo limit our analysis to plants that re preset ring all years ofthe analysis, (For Colombia, the sample period is 1981 through 1991, for Mexico itis 1986 ‘trough 1990 and for Morocco itis 1986 through 199.) Exclusion of entering and ‘ting plants obviously opens the door to selection bis in our ndings, butt
substantially simplifies the econometric modeling, We fee ths price worth paying because the omited firms supply avery small faction a the matket More importantly, ‘were les concemed wih precise parameter estimation than with simply asking whether significant pattems of association are present It is highly unlikely that they would be
Trang 24‘A Performance determinants
‘Our results foreach 4-igit chemicals industry in Colombia, Morocco and Mexico are presented in tubles A2,1 trough 2.3 of appendix 2, respectively (Industries with fewer than 10 plants continually present are not reste.) Parameters of the covariance _matrces forthe compound disturbances are not reported to conserve space, but we do
‘report the coefficients from equation (8) tat relate the unobserved plant effets, "and 10 inital realizations on the performance variables, a, and ¢,
Under each industry heading the left-hand column reports parameter estimates and the right-hand column report the associated standard erors Wa tet statistics for the ml hypotheses that product quality is unrelated to international activities and marginal cost are unrelated to international activites are reported near the bottom of cach able, All coefficient estimates that are at least twice their standard ero are reported in bold as are z” statistics with p-values less than 0.05 The degrees of freedom forthe
‘Wald statisties depend upon the number of activity bundles thet are considered, which in tum vary across countries and industries (The ater occurs within a country because some industries donot exhibit ll posible activites.) Parameters, standard erors and test statistics that desribe the relationship between activities and performance are reported in shaded panels,
Demand parameters
Parameter estimates forthe demand system appear inthe top pane! of each table [Note thatthe parameter that measures the sensitivity of indirect uly to pies always quite significant and postive, as hoped This gives us some confidence thatthe
Trang 25consumers Is generally higher in Moroco than nthe eter countries because the French accounting system thee led toa unit pice measure ther that was somewhat lower than the measure we calculated for Mexico and Colombia; hence the mad assigned that country higher demand elasticities
Estimates of are also quite accurate They imply that for most industries, the standard eror of ¢ + is roughly twice he standard err of alone Or for given
‘consumer, indirect ules vary subtatally across regions, as well as between foreign and domestic varieties (t would be straightforward to use these igure to calculate measures of ewn-rgion or home marke bia, but we have not yet done so)
Finally, y measures the effet ofan increas in the price of imported goods onthe rlatve utility attained from home goods This parameters no estimated as accurately ax the other demand parameters because itis identified solely by temporal variation, andthe
umber of years we observes limited Nonetheless, it is positive in cight of the nine cases where itis statistically significant atthe an level (Fhe exception ithe detergent industry in Colombia.) Overall, then, our estimates of he demand system conform very nicely to priors
Product quality determinants
“The next panel in ables AI through AL3 reports estimates ofthe parameters ‘hat appear in equation (a Before considering the parsmtesof primary intrest let us
recap our results onthe control variables First, conditioning on lagged performance and intemationl activites, most industries we analy in Colombia nd Morocco showed
no significant trend in relative quality The except 1s were Colombian detergents and Moroccan detergents and pants, each of which exhibited a tendency to improve relative
Trang 26to imports over the sample period Mexico, in contrast, hosts a number of ndustes that fell increasing behind imported subsites during the sample years, Rubber products and ‘Pharmaceuticals tended to improve but five ofthe six remaining industes showed significant negative trends
(One might expect that plants begining the sample period with large capital stocks would exhibit relatively high quality, since iil capital stocks should reflect pre- sample demand for their products Indeed, nal ofthe industries wher intial capital stocks proved statistically significant, they were positively comelated with relative ‘product quality This relationship might sem spurious, since product quality is related 10 marketshare by the identity (6), and firms with large market shares surely have large capital socks However, consumers are only about prices and product appeal, not productive capacity So, ifour mode! is comeetly specified and prices are property measured, any relation between relative quality and inital capital stocks is indeed a ‘consequence of interaction beeen quality and size
‘The next conrl variables are lagged quality and lagged marginal cost The formers significant in almost al ass, implying that our quality measure follows an autoregressive process, conditioned on othe regressors Given that product
characteristics and reputation evolve slowly over time, this is what one would expect to find In most cases the AR(1) coefficient is significantly les than unity, but Colombian pharmaceuticals and Mexican ferilizerspesticides yield roots close to one (We have not tempted unit oot tests for ou system.) Coefficients on lagged marpnel cost are usally
unimportant, 0 for most industries marginal cost shocks have litle effect onthe
Trang 27‘marginal cos is significant, firms adjust to cost shocks partly by reducing the future quality oftheir product
Finally, the coefficients om initial qu ty realizations indicate that unobserved heterogencity is often important, and that Wooldidge's correction matters, On the other hand, coefficients on inital marginal cost realizations are usualy insignificant, so our ‘representation of unobserved heterogeneity (6) is probably more general that it needs to be for most industries Nonetheless, the covariance matrices for our compound
<istubances imply that in many cases, persistent unobserved heterogeneity remains afler conditioning on these variables
‘Consider now the variables of primary intrest—dummies forthe various atvity ‘bundles Wald statistics reported atthe bottom ofeach able test the joint null
‘nypothess that ast period's activity bundles have nothing to do with current quality Tellingly, all but three ofthese 7 statistics have a p-value greater than 0.05, so in 16 of ‘the 19 countryZindustry cases, one cannot reject the null hypothesis that previous
imternational activities have no effect on current product quality Experiments with additional lags on activity bundles (not reported) lft his basic message intact,
Further, in the thre instances where activites are significantly related to quality, ‘the dominant relationship is negative In Moroccan plastics and Colombian
Trang 28
4o beter: (In this case none ofthe coefficients for hese activity bundle is individually significant, so colineadty problems prevent us fom drawing strong conclusions.) (Overall, then, our methodology yield virtually no evidence that intemational activities Granger cause product quality improvements (We shal return tothe issue of whether we
are missing significant linkages shor.)
Marginal cot determinants
‘The weak and occasionally negative association between product quality and intermationl activities need not imply tat Sms pursuing these activities are misguided, It may be that international activites mainly help to reduce costs rather than improve quality Are such inkages picked up by our model?
Parameter estimates for out marginal cos equation (Ib) are reported by industry and country inthe lower panels of tables A through AL Again we begin our isoussion of the results withthe contol variable First note that most industries show
To significant trend in marginal coats, once initia capital stocks lagged costs and activities ae controlled fr The exceptions ae plastic and soap in both Mexico and Morocco, which trend significantly upward, and both fertilizers in Mexico and ‘Pharmaceuticals in Colombia, which tend downward
“More interestingly, productive capacity (measured by inital capital socks) seems to have litle todo with our measure of marginal production costs Except among
Mexican robber producers, where marginal oss ise with capacity and Mexican
‘pharmaceutical producers, where the opposite occurs, initial capacity is not «significant
‘We use the following abbreviations: X (export, Ml (npr of nemeie ood), MK (inprer feel gods), nd FDI (i what les ive pect regs owner
Trang 29predictor of our cost measure, Tis suggests that scale economies are not dramatic and that production technologies are reasonably homothetic over the size range we observe in our panels
As forthe role of lagged marginal costs, all ofthe industries show clear evidence of serial dependence Not surprisingly, the coofficients are universally greater than eo, and most are at least two standard deviations below unity In half ofthe Mexican
industes, lagged quality i also associated with current costs, usually positively, but no such relationship emerges in the other countries Finally, intial realizations on marginal ‘costs are significant in most sectors, implying that unobserved heterogeneity is important
‘The (unreported) variance in residual plant effects is aso often significant, and correlated ‘with the residual plant effect in the quality equation, However, the sign of this correlation ‘varies fom industry to industry
"Now consider the relation between international activities and marginal costs Except in Morocco, our Wald tess flo reject the mull hypothesis thatthe two are unrelated There, FDI alone is associated with significant cost redutions among
pain varnish producers, but exports and FDI together lead fo cost neteases Interestingly, ‘he plans that both export and have FDI are also the ones predicted (weakly) to have the highest quality product, so the combined effet ofthese activities may well be to enhance profits" Exporting and FDI tend to increase costs among detergentperfume producers 100, but here there is no evidence of offsetting quality gains, Finally, among Moroccan
pharmaceutical producers, exporting and PDI significantly reduce future costs, alone or "5 nissaighforaré to aut the predicted fet ofthese marina ost and quality changes on ofits, fm Sm, bat me Rave na Set done 8
Trang 30together They also tend to reduce quality, however, so as with Moroccan paint/vamish, the net effect ofthese activities on profits is n
diately apparent,
Although none of the ther country/industry panels shows a significant joint relationship between international atvies and future marginal cost, there are instances ‘where particular activity bundles yield statistically significant effects These occur ‘exclusively among bundles that involve importing intermediate inputs, In Colombia,
ferilizeripesticide producers reduce thee future costs by importing intermediates, but ‘they reduce the quality of thei product too In Mexico, intermediate imports increase
‘marginal costs but also increase quality among rubber producers The same is true among Mexican ferilizerfpesticide produces, although the quality effect is nt statistically significant Further, among Mexican pharmaceutical producers, imported
{ntermediates—in combination with exports or imported capital goods—reduce marginal costs and, if anything, tend to increase product quality Only inthe “other chemicals” ‘sector do imported imported intermediates (in combination with imported capital and ‘exports significantly increase future costs without also tending to improve product quai
‘Overall, then, most country/industry panels show no significant association ‘between international activities and future marginal costs However, a minority of
industries exhibits this type of causal link Further, whether the activities inerease or reduce costs ther is often an offctting change in product quality, making the net effect ofthe intemational activites on profits ambiguous
Trang 31A robustness check
‘We mentioned earlier that Granger causality tests may miss important linkages if the transmission of knowledge does not actually coincide with the observed international activity, For example, when a foreign buyer places an order with a domestic firm, it may ‘wansmit blueprints and technical esistance year before the actual exports occur To
check whether this problem has undermined out inferences, we now abandon or dynamic specification in favor of a model that simply tests for static correlation between
temational activities and performance Tha is, we drop alla,., and c,.g variables and
‘we replace by, with b, on he right-hand side of equations (La) and (1b) Isis model reveals no association, we have stronger evidence against the claim that international activities transmit knowledge, On the other hand, statistical significance need not imply causal relationships from activities to performance, It may reflet causation from performance to ativities (which we shall explore shorty), or it may reflect transitory effects that do not involve knowledge transmission, For example, as Ethier (1982) suggested, access to imported intermediates may simpy enrich the menu of inputs for firms and allow them to produce beter products or reduce their cost
‘Tables A1.4 through A1.6repor the static version of our model for each industy7eountry panel (We stil use an eror component specification, but we drop ‘Wooldtidge's correction fr inital conditions because lagged performance measures no Jonger appear on the right-hand side.) Notice first cat, when lagged quality is dropped nd international activities are no longer lagged, many sectors show a significant
association between product quality and intemational ctivty Further, the association is ‘now positive in Colombia for firms that import their intermediete inputs By itself, his
Trang 32result has several interpretations It might mean that importing intermediates allows a fim to lear something about technology and to permanent increase the quality of product, 8 Grossman and Helpman (1991) suggest Or it might reflect he static benefits fom importing that Ethier (1982) wrote of The Iter interpretation strikes us a ‘more plausible, given that we picked up no evidence of Granger causation fom imported intermediates to subsequent product quality guns, and that imported intermediates are unlikely to be preceeded by ongoing contacts with foreign suppliers
Morocco and Mexico also exhibit mor significant associations in the static ‘model However, in contrast to Colombia international ecivitis stil sem to hurt «quality, ifanything, Thus, in determining the effects of international activites on performance, country conditions may be more important than technological features of indusies that are common across borders
Finally, inthe marginal cost equation, coefficients differ from those inthe <éynamic model but the evidence tat international activities are associated with lower ost i, if anything, weaker In sum, with the exception ofthe Colombian resus on ‘product quality, te timing ise does not appear tobe the main reason that firms" product
‘quality and marginal costs are unrelated to their international activities BL Activity determinants
‘Our final empirical exercise addresses the issues of how activities ae chosen, and ‘whether performance cases international atvities, We begin with a descriptive review
16 = tipors may embody iene intermediates th eno avalble nthe local eueomy ‘The peter the quant of ch import, he reser eras il be he pber sigh at al researchers gin fom inspecting and sig thse goods" (Grssman and Helpean, 1991 168)
Trang 33‘ofthe activity pattems found in our data ses Without controling for anythin, tables ‘21 and A22 ropor probabilities of diffrent atvity bundles and transitions between bundles based on simple cll counts Clery, rgardes of which atvty combinations we consider, the pursuit of one activity increases the likelibood that others wl be pursued aswell imilar statements hod foe transitions Ione ofthe interatonal
tivities has ben init, the probability tht others wil follow increases Accordingly, pirical models that focus ona single activity a the key to inteational technology iffosion probably suffer from significant omited variable bias
‘Ofcourse, pattems of association may not reflect complementarties among the sctiviis, is possible that certain frm characterises location ina por ity or
ethnic ies to foreign counties may make them engage in multiple international activities even when tie complementares are absent To gett this possibility, and to biter understand firms" activity choices, we next estimate the system of linear
probability equation (2), allowing for unobserved heterogeneity and using 'ooldddees (2000) correction forthe initial conditions problem
Table A2.3 presents els for Colombia We report entimats for hee ofthe four possible atvty bndles—the bundle with no international etivities the omitted
«category (Parameter estimates fortis option canbe derived from the consent thatthe sun ofthe probabilities ofthe diferent bundles must lays be unity) The fist column reports results forall firms, but des ot include performance measures Two points merit, mention First activites are highly persistent, even afer controling fr unobserved heterogencty This suggests that the start-up costs associated withthe initiation of new
Trang 34have lasting effects on activity pattems (Roberts and Tybout, 1997) Second, we continue to find thatthe probability that a firm becomes an exporter is higher if was already porting intermediates inthe previous year, and the probability of becoming an importer jis higher for exporters than for nonexporters However, fst reported at the bottom of | ‘able A3.1 fil to reject the nul hat these differences in conditional probabilities are 2210 Thus, although imported intermediates and exporting tend to go together, the <éynamic interactions between them are not strong in tis fller specification
‘Our second st of estimates (columns 3 and 4) adds performance measures to investigate whether firms with high quality and/or low costs are more likely to engage in activities Since our quality variable is normalized to have the same mean in each of the ‘our-digi industries, we cannot simply pool al sectors Instead, we restit attention to the largest four-digit industry in tems ofthe numberof firms—pharmaceuticals” ‘Within this four-digit industry, the two patterns noted above continve to sppear
Activities are highly persistent, and the presence of one activity increases the probability ‘thatthe other will be initited However, these dynamic interactions are not statistically significant,
Regarding the effets of quali
and cost onthe choice of activities, the signs of ‘the coefficients ar intuitive forthe “both export and import” bundle Firms with higher ” Ahensirsy we might hve pooled al fms a aed industry dummies The problem with his ‘approach shat as age mmbe of atonal ureters tobe estimated, ad nce several of
‘Meter industries have gute sal populations of ens his opi exists dees of feedom A dawbsck of focusing on the gest industry however ta the en summuvaag ompesionperfermance exhibis very le cưe-fơm vara, ad so we drop thi fom the ‘stmt rally, eve fr sited specfeton he numer of parameter 1 be eximated it uc kược.KxGF.1) elements inthe vector of coon on aged ates 8, 2=KxC2™1) ‘ements ia vector ofcoetcent os lagged peTomance andthe th sme nme gin er the {nial condone cecton For hs aso we ae unable etnae he model for Mexico, whe
Trang 35
auality and lower costs are more likely to pursue this bundle of activites, However, neither cocticient significant, nor are the cocfcient for theater bundles, so the case for causation fom previous performance to activity choices is weak Table A2.4 reports analogous results for Moroco, where we observe data on exporting and foreign equity
holdings The results ae broadly similar to those in Colombia Activites are persistent, however, the degree of persistence as measured bythe magnitude ofthe coefficients on own lagged activities is penerally smaller, This is expecially te for FI, where past DI activities raises the probability of eurrent FD ativity hy nly 10 percent inthe ist setof regressions, and not tal in the second st including peformance variables (which
are reported only for plastics)
(One puzzling exception othe patter of insignificant eects of performance on the choice ofaetvitcs is in the equation forthe choice of both sctiviics, Other things sual, Sms with higher quality and lower costs are ls likely to choose both activities “This mors the negative correlation between product quality and joint pursuit of exports snd FDI found in tables AL2 and ALS Taken at fee value, it suggest that foreign
investors ar attracted to plants that have been performing relatively poorly
1 Summary
“This paper has several basic mestags Fest, by imposing enough stucture on the ‘modteton function andthe demand system, itis possible to measure product quality and marginal cost atthe plant level and to late the evolution ofthese variables to firms’
activity histories Doing so, we find strong irm-level persistence in both quality and
Trang 36
rarpinal costs, as expected, However, in most industryfcountey panels we studied, past intemationl activities do not belp much to predict curent performance, once past, realizations on quality and marginal cost are controlled for, Thats, activites donot ‘ypically Granger cause performance Interestingly, in the minority of cases where significant associations emerge, ntemationalstvites appear to move ost and product quality in the same direction, Thus, the net effect on profits in these cassis not
imedisely apparent
Concerning the determinants of international activities, several basic pattems «merge Mos fundamentally, activites are highly persistent, even aftr one controls for ‘unobserved heterogeneity This suggests that fms incur sunk threshold costs when they initiator cease activities, o temporary policy or macro shocks may have log run fects onthe pattems of activities observed in a particular country o industry
‘Second activities tend to go together Thus, studies tat eae firms’ performance {to on interatonal activity and jgnore the otbers may generate very misleading
‘conclusions On the oter hand, the bundling of activities together ses to manly reflect “unobserved plant characterises, lke managerial philosophy, contacts, product niche and locaton, One these are controle for, there site evidence that engaging in one interational activity increases the probabil tht others wil occur in the future
Finally, we find weak evidence fom Colombia tha firms that ar already song performers seek out foreign markets as away to enance ther profits, But these results do not generalize to Morocco, where plastics fim tat have been performing poodly are
Trang 37Bibliography
Ackersburg, Daniel and Mark Rysman (2001) “Unoserved Product Differentiation in <iscrete Choive Models: Estimating Price Elasticities and Welfare Effects,” processed, Boston University
‘Anderson, Simoa, Andre de Palma and Jacques Francois Tisse (1992) Discrete Choice Theory of Product Differentiation Cambridge: MIT Pres Aitken, Brian and Ann Harison (1999) "Do Domestic Finns Benefit fom Foreign Direct Investment? Evidence from Panel Data,” American Esonomic Review R9); 605-
618
‘Av, Bee-Van and Amy Hwang (1995) “Productivity and the Export Market: A Fiem- Level Analysis,” Journal of Development Economics 47: 313-332, Berry, Seven (1994) “Estimating Diserete-Choice Models of Produet Diferetiation;” ‘Rand Journal 25(2), pp 242-262 ‘Berry, Steven, James Levinsohn and Ariel Pakes (1995) “Automobile Prices in Market Equilibrium,” Beonometrica 63), pp 841-890, Bigsten, Ame, Paul Collier, Stefan Dereon, Bernard Gauthier, Jan Gunning, Anders Isaksson, Abena Oduro, Remico Ooostendorp, Cathy Pall, Mans Soderbom,
“Michel Sylvain Prancis Teal and Albert Zeufack (1999) “Exports and Fiem-Level Efficiency inthe African Manufscturing Sector,” Sehool of Busines, U Montreal
Blomstrom, Magnus, Robert Lipsey and Mario Zejan (1994) “What Explains the Growth ‘of Developing Counties?” in Baumol, Wiliam J, Richard Nelson and Edward Wolf, eds, Convergence of productivity: Cross-national Studies and Historical Evidence Oxford and New York: Oxford Univesity Press, pp 243-59,
(Che, Tain, and De-piao Tang (1987) “Comparing Technical Efficiency Between Import Substituting and Export-Oriented Foreign Firms in a Developing Country,” Jounal of Development Economics 36; 277-89
(Coe, David and Ethanan Helpman (1995) Alnternational R&D Spillovers,¢ European Economic Review 39, pp 859-887
Trang 38
Djaskov, Simeon and Berard Hoekman (2000) “Foreign Investment and Productivity ‘Growth in Czach Enterprises," World Bank Economic Review 14(1), p 49.64
de Long, J Bradford and Lawrence Summers (1991) “Equipment Investment and Long ‘Run Grovith,” Quarterly Journal of Economics 106(2), pp 445-502 Eaton, Jonathan and Samuel Kortum (1996) “Trade in Ideas: Patentng and Productivity in he OBCD,” Journal of International Economics 40 pp 251-278, Ethier, Wilfred (1982) “National and International Retums to Scale inthe Moder ‘Theory of Intemational Trade,’ American Economic Review 72(3) 389-405, Ericson, Richard and Ariel Pakes (1995) “Markov-Perfet Industry Dynamics: A Framework for Empirical Wotk,”Review of Economic Stadies 62, pp, 3-82, Geweke, John, Michael Keane and David Runkle (1997) “Statistical Inference in the Multinomial Mult-Perod Probit Model,” Journal of Econometrics 80, pp 125-
165,
Goldberg, Pinelopi and Michael Kneter (1997) “Goods Prices and Exchange Rates: ‘What Have We Learned?” Journal of Economic Literature 35(3), pp 1243-1272 (Grossman, Gene and Ethanan Helpman (1991) Innovation and Growth inthe Global ‘Economy Cambridge, MA: MIT Press Hiaddad, Mona and Ann Harrison (1993) “Are There Positive Spillovers from Direct Foreign Investment? Evidence from Panel Data for Morocco," Jounal of
Development Economics 82: 51-14,
Hobday, Michael (1995) Innovation in East Asia: The Challenge to Japan Cheltenham, UK: Edward Elgar Katz, Jorge, et l (1987) Technology Generation in Latin American Manufaccuring “Industries, London: MacMillan ‘Keane, Michael (1992) “A Note on Identification ofthe Multinomial Probit Model,” Journal of Business and Economic Statistics 102), p 193-200 Keller, Wolfgang (1998) AAre international R&D Spillovers Trade-elated? Analyzing Spillovers Among Randomly Matched Trade Partners,” European Economic
‘Review 1998: 1469-1481
Keller, Wolfgang (2000) “Do Trade Patterns and Technology Flows Affeet Productivity ‘Growih?” World Bank Economic Review 214(1, pp 17-7 kim, L (1977) Imitation to Imovation: The Dynamics ‘Cambedge, MA: Harvard Business School Press of Korea's Technological Learning,
Trang 39‘Kraay, Aar (1909) “Expors and Economie Perfortmance: Evidenee fiom CChinese Enterprises,” Revue d' économie du développement, 1-2, pp.183-207 a Pansl of Knetter, Michael and Penelopi Goldberg (1999) “Measuring in Export Markets,” Journal of Inernational Economics 41(1), pp, 27-60 the Intensity of Competition Lall, Sanjaya (1987) Learning to Indusrialize London: MacMillan, 1987,
Lu, Shihua and James Tybout (2000) “Inferring Plnt-Level Performance Measures from Revenues and Cots, proceed, Pennsylvania Slate University ‘MeFaden, Daniel (1974) “Conditional Logit Analysis of Qualitative Choice Behavior,” in Paul Zarembka, ed, Fromirs in Economerics, New York: Academic Press,
1974,
“Milgrom, Paul and John Roberts (1990) * The Economies of Modern Manufacturing: ‘Technology, Strategy and Organization,” American Economic Review, 80(3), Pp 311-588
Pack, Howard (1987) Productivity, Technology and Industrial Development New York: (Oxford University Press Pack, Howard (1999) "Modes of Technology Transfer atthe Firm Level," processed, ‘Wharton Schol, University of Pennsylvania Rhee, Yung Whee, Bruce Ross-Larson and Garry Pursell (1984) Korea's Competitive Edge: Managing Enry ito World Markers Balinore: Johns Hopkins Press
Roberts, Mark and James Tybout (1997), “The Decision to Export in Colombia: An “Empirical Model of Entry with Sunk Costs," American Economic Review 87, pp 545-64 “Tybout James (2000) “Manufacturing Firms in Developing Countries: How Well 6o “They do, and Why?” Journal of Economic literature 38(1, pp 11-44
Tybout, James and M Daniel Westbrook (1995), "rade Liberalization and Dimensions of Eiieney Change in Mexican Manufacturing Industries," Journal of International eonomics 38, p S378
‘Westphal, Larry (2001) “Technology Strategies for Economic Development ina Past ‘Changing Global Economy,” forthcoming in Economics of novation and New Technology
‘Wooldridge, Jeffrey (2000) “The Initial Conditions Problem in Dynamic, Nonlinear ‘Pane! Data Models wth Unobserved Heterogeneity,” unpublished working paper, Department of Economies, Michigan Sate University