manufacturing firms in developing countries:how well do

Journal of Economic Literature Vol. XXXVIII (March 2000) pp.11–44

Manufacturing Firms in DevelopingCountries: How Well Do They Do,

and Why?

JAMES R. TYBOUT1

1. Overview

THE MANUFACTURING SECTOR isoften the darling of policy makers in

less developed countries (LDCs). It isviewed as the leading edge of modern-ization and skilled job creation, as well asa fundamental source of various positivespillovers. Accordingly, although manyLDCs have scaled back trade barriersover the past twenty years, the industrialsector remains relatively protected in thetypical country (Maurice Schiff and Al-berto Valdez 1992, ch. 2; Refik Erzan etal. 1989; Francis Ng 1996).2 Governmentsalso promote manufacturing with specialtax concessions and relatively low tariffrates for importers of manufacturingmachinery and equipment.

At the same time, many observers be-lieve that the maze of business regula-

tions is unusually dense and unpre-dictable in LDCs.3 Summarizing an ex-tensive survey of managerial attitudesaround the world, Aymo Brunetti, Greg-ory Kisunko, and Beatrice Weder(1997) report that LDC firms generallyconsider the institutional obstacles todoing business more burdensome thantheir OECD counterparts.4 The regula-tory problems that they view as moresevere include price controls, regula-tions on foreign trade, foreign currencyregulations, tax regulations and/or hightaxes, policy instability, and generaluncertainty regarding the costs of regu-lation. Other types of regulation—in-cluding business licensing and laborlaws—are not viewed as especially bur-densome on average in the LDCs, butconstitute major problems in certaindeveloping countries.5

11

1 Pennsylvania State University. This paper waspartially funded by the World Bank’s EconomicDevelopment Institute. I have benefited from dis-cussions with Bee-Yan Aw, Mark Dutz, Mark Ger-sovitz, Bernard Hoekman, Carl Liedholm, WilliamMaloney, Desmond McCarthy, Mark Roberts,Kamal Saggi, Maurice Schiff, and William Steel. Iam especially indebted to John Pencavel, JohnMcMillan, Howard Pack, and several anonymousreviewers, who have improved this manuscriptwith many helpful suggestions.

2 The need for revenue is a second motivationfor relatively high tariffs in developing countries,although nontariff barriers seldom serve this func-tion.

3 A well-known example of the problem wasgenerated by the Institute for Liberty and Democ-racy in Peru, which attempted to register a ficti-tious clothing factory in the mid-1980s. “To regis-ter the imaginary factory took 289 days andrequired the full-time labor of the group assignedto the task, as well as . . . the equivalent of 23minimum monthly wages” (Hernando de Soto1989, p. xiv).

4 For purposes of this paper I will ignore thefact that Turkey, Mexico, and Korea are membersof the OECD.

5 There were no major differences between theLDCs and the OECD in terms of the regulations

Moreover, within the manufacturingsector, it is also often argued that poli-cies favor large firms while inhibitinggrowth among small firms (Ian Little1987). In some cases, investment incen-tives are available only to projectsabove a minimum scale, and large-scaleproducers are singled out for specialsubsidies.6 Anti-trust enforcement istypically weak, and special tax breaksare sometimes meted out to large, influ-ential corporations (Bernard Gauthierand Mark Gersovitz 1997).

Even when policies do not explicitlyfavor large firms, these firms may enjoyde facto advantages. Banks view them asrelatively low risk and cheap to service(per unit of funds lent), so they havepreferential access to credit.7 This phe-nomenon is relatively marked in devel-oping countries because private sectorcredit is relatively scarce there, infor-mation networks are poorly developed,and binding interest controls are rela-tively common (Ross Levine 1997; Little1987; Tybout 1984).

Protectionist trade regimes are alsomore likely to favor large firms, bothbecause these firms’ products competemore directly with imports, and because

sectors with large, capital-intensivefirms lobby the government more effec-tively. Further, while many of the costsof dealing with dense regulatory re-gimes are fixed, the payoffs from doingso probably increase with the scale ofoperations. For example, access to thelegal system, access to the formal bank-ing sector, and publicly administeredemployee benefits are relatively valu-able to large producers (Alec Levensonand William Maloney 1997). Hence,moderate-sized firms, which are largeenough to show up on regulators’ radarscreens but too small to gain much fromcompliance, may be punished the most.

These basic tendencies of LDCs—toward industrial sector promotion,dense, unpredictable regulatory re-gimes, and credit markets or commer-cial policies that favor large firms—raise a number of fundamentalempirical issues. First, do the regula-tory regimes and the bias against smallproducers prevent small firms fromgrowing, and thereby create losses dueto unexploited scale economies? Sec-ond, if these regimes prevent smallfirms from threatening the larger in-cumbents, do LDC industrial sectorslack dynamism and competition? Thatis, have entrenched oligopolies emergedthat are neither innovative, technicallyefficient, nor likely to price competi-tively? Finally, has trade protectioncompounded the technical inefficien-cies and monopoly power that arisefrom regulatory regimes? In this paperI selectively take stock of what we havelearned about these issues from firm-and plant-level data sets over the pasttwenty years.

I shall begin by briefly reviewingsome of the distinctive features of theenvironment in which LDC manufac-turers operate. This will serve as back-ground for the discussion that follows,and help to distinguish those features of

concerning new business start-ups or safety andenvironmental standards; further, LDC firmsviewed labor regulations as less of a problem thandid OECD firms. But licensing and labor lawshave been flagged as major problems in India andsome Latin American countries, inter alia. Forcountry-specific discussions of the regulatory bur-den, see de Soto (1989) on Peru; World Bank(1995a) and Tyler Biggs and Pradeep Srivastava(1996) on sub-Saharan Africa; Little, DepakMazumdar, and John Page (1987) and Garry Pur-sell (1990) on India. Severance laws in developingcountries are discussed in World Bank (1995b,Chap. 4), and in more detail for the Latin Ameri-can case in Alejandra Cox-Edwards (1993).

6 See Howard Pack and Larry Westphal (1986)on Korea; Mariluz Cortes, Albert Berry, and Ash-faq Ishaq (1987) on Colombia; Robert Wade(1990) and Bruch and Hiemenz (1984) on E. Asia.India is an exception—see, for example, Little,Mazumdar, and Page (1987).

7 Levine (1997) provides references.

12 Journal of Economic Literature, Vol. XXXVIII (March 2000)

LDC manufacturers that trace to struc-tural characteristics of their economiesfrom those induced by industrial, trade,or labor policies. Next, drawing on theavailable evidence, I will take up the is-sue of whether small firms have beensomehow suppressed, and more gener-ally, whether the LDC business envi-ronment has bred noncompetitive pricingbehavior and low productivity. Finally, Iwill address the question of how tradeprotection has conditioned pricing,efficiency, and productivity growth.

2. The Business Environment: What’s Different in LDCs?

In terms of income levels and busi-ness environment, the countries typi-cally labeled “developing” are a very

heterogeneous group. By the WorldBank’s reckoning they currently span theper capita income range from $US 80(Mozambique) to $US 8,380 (Argen-tina).8 Nonetheless, looking acrosscountries, some distinctive features ofthe business environment become in-creasingly evident as one moves downthe per capita income scale. At the riskof over-simplifying, I will begin by men-tioning the most striking and widelyacknowledged among them.

Market size. Although some develop-ing economies are quite large, most arenot. Hence, excepting countries like

8 This income range describes all countries thatare not classified by the World Bank as high in-come or transition. Figures are taken from theWorld Bank’s World Development Indicators1998.

Tybout: Manufacturing Firms in Developing Countries 13

Brazil, China, India, and Indonesia, thesize of the domestic market for manu-factured products is relatively limited(Figure 1). Further, among the least de-veloped countries, Engel effects favorbasic subsistence needs over all butthe most basic manufactured products(Figure 2). So when transport costsare significant and the OECD countriesare distant, demand for the moresophisticated manufactured goods issmall.

Access to manufactured inputs. Themenu of domestically produced inter-mediate inputs and capital equipment isalso often limited in developing coun-tries. Thus producers who might easilyhave acquired specialized inputs if theywere operating in an OECD countrymust either make do with imperfect

substitutes or import the needed inputsat extra expense. This latter option isthe dominant choice among smallercountries.9

Human capital. Low rates of secondary9 As a crude exercise, one can designate “non-

electrical equipment,” “electrical machinery” and“transport equipment” as the machinery andequipment industries. Then using domestic pro-duction data from the World Bank and trade datafrom the COMTRADE system, the ratio of net im-ports to domestic consumption for this group ofproducts can be constructed product by product.For the 53 countries with complete data, 75 per-cent of the cross-country variation in this measureis explained by the logarithm of GDP and adummy for transition economies. The coefficienton the logarithm of GDP is –0.14 and it carries at-ratio of 9.69. (Surprisingly, the log of GDP percapita adds no additional explanatory power to theregression.) The predicted share of imports in do-mestic consumption of machinery and equipmentis greater than 0.6 for economies with GDPs lessthan 12.5 billion 1987 US dollars.


education and a scarcity of techniciansand scientists also affect the mix ofgoods manufactured and the factor pro-portions used to produce them.10 Simi-larly, many have argued that flexibilityin production processes and the abilityto absorb new technologies are directlyrelated to the stock of domestic humancapital (e.g., Richard Nelson and Ed-mund Phelps 1966; Robert Evenson andWestphal 1995; Wolfgang Keller 1996).

Infrastructure. Roads, ports, airports,communication facilities, power, andsafe water access tend to be relativelylimited in LDCs (World Bank 1994, p. 3,Figure 1; p. 13, Table 1.1, and p. 14,Figure 1.1). Production techniques aredirectly affected, as are the costs ofservicing distant markets. Poor trans-portation networks are particularly lim-iting in the least developed, more agrar-ian economies, where consumers arespread throughout the countryside. Ininstances where infrastructure servicesare missing or unreliable, some firms mustproduce their own power, transport,and/or communication services.

Volatility. Macroeconomic and rela-tive price volatility is typically more ex-treme in developing countries. Histori-cally, Latin America and sub-SaharanAfrica have stood out as the most vola-tile, but all developing regions havedone worse than the industrializedcountries (World Bank 1993; RicardoHausmann and Micheal Gavin 1996).

Governance. Finally, legal systemsand crime prevention are also relativelypoor in developing countries, and cor-

ruption is often a serious problem(World Bank 1997; Brunetti, Kisunko,and Weder 1997). Hence the protection ofproperty rights and contract enforcementcan be problematic.

3. Plant Sizes and Scale Efficiency

Combined with industrial sector poli-cies, the above circumstances (and oth-ers I have neglected to mention) lead toseveral distinctive features of LDCmanufacturing sectors. Perhaps themost striking of these is their dualism.In many industries, large numbers ofmicroenterprises and a handful of mod-ern, large-scale factories produce simi-lar products side by side.11 The smallproducers frequently operate partly orwholly outside the realm of governmentregulation, and rely heavily on informalcredit markets and internal funds for fi-nance. They are relatively labor inten-sive, so they account for a larger shareof employment than of output.

3.1 The Size Distribution

The contrast between the size distri-bution of plants in developing countriesand that found in the OECD is dra-matic. Table 1 provides some crudecomparisons. Note that there is a largespike in the size distribution for the sizeclass 1–4 workers, and it drops offquickly in the 10–49 category amongthe poorest countries.12 This is not true

10 The wages of scientists and engineers inmanufacturing firms constitute 0.2 percent ofGDP in the most technologically primitive of thedeveloping countries, while they account for 1.0percent of GDP in the OECD (Evenson andWestphal 1995, Table 37.1). A logarithmic regres-sion of the secondary school enrollment rate onGDP per capita yields an elasticity of 0.62 and anR-squared of 0.65. (Calculations are based on theBarro-Lee data base, available at www.worldbank.org/growth/ddbarle2.htm.)

11 In Colombia, for example, more than half ofthe 5–digit ISIC industrial sectors that containplants with less than ten workers also containplants with more than two hundred workers. Ibase this statement on Colombia’s annual manu-facturing survey, which neglects most plants withless than five workers, so this statistic substantiallyunderstates the prevalence of microenterprises.

12 Many microenterprises are invisible to officialcensus takers because they do not have postalboxes, are impermanent, and/or are part of farmcompounds. Thus Table 1 substantially under-states their prevalence. “[C]omparison of villageby village enterprise censuses conducted by[Michigan State University] and local scholars with


in the United States or other industrial-ized countries. The emphasis on small-scale production not only correlatesnegatively with per capita income levelsacross countries (Carl Liedholm andDonald Mead 1987, p. 16; RanadevBanerji 1978), but also within countriesthrough time (Ian Little, Dipak Mazum-dar, and John M. Page, Jr. 1987; WilliamSteel 1993).

What accounts for this phenomenon?For some LDCs, observers have

pointed to regulations and taxes that areenforced only among the large, formalsector firms. James Rauch (1991) for-malizes this explanation by extendingRobert Lucas’s (1978) model of thefirm size distribution with heterogene-ous worker/entrepreneurs. He showsthat when larger firms face higher unitinput costs, the most talented entrepre-neurs operate big firms to exploit theirproductivity advantage, and the extraprofits they earn from being big morethan cover the higher input costs theymust pay. Less talented entrepreneurs staysmall and informal. The size distributionexhibits a “missing middle” because it

‘official’ censuses shows that the latter not infre-quently undercounted the number of enterprisesby a factor of two or more” (Carl Leidholm andDonald Mead 1987, p. 20).

TABLE 1DISTRIBUTION OF EMPLOYMENT SHARES ACROSS PLANT SIZES

Numbers of Workers

1–4 5–9 10–19 20–49 50–99 >99

United States, 1992a 1.3 2.6 4.6 10.4 11.6 69.4Mexico, 1993b 13.8 4.5 5.0 8.6 9.0 59.1Indonesia, 1986c 44.2 17.3 38.5S. Korea, 1973d 7.9 22.0 70.1S. Korea, 1988e 12 27 61Taiwan, 1971c 29.1 70.8Taiwan, 1986f 20 29 51India, 1971g 42 20 38Tanzania, 1967g 56 7 37Ghana, 1970g 84 1 15Kenya, 1969g 49 10 41Sierra Leone, 1974g 90 5 5Indonesia, 1977g 77 7 16Zambia, 1985g 83 1 16Honduras, 1979g 68 8 24Thailand, 1978g 58 11 31Philippines, 1974g 66 5 29Nigeria, 1972g 59 26 15Jamaica, 1978g 35 16 49Colombia, 1973g 52 13 35Korea, 1975g 40 7 53

Sources:a 1992 U.S. Census of Manufacturing, unpublished Census Bureau calculations.b INEGI (1995).c Steel (1993).d Little, Mazumdar, and Page (1987, Table 6.5).e 1988 Census of Manufacturing, Republic of Korea, calculations of Bee-Yan Aw.f Chen (1997, Table 2.2).g Liedholm and Mead (1987).


never pays to be just large enough toattract enforcement. Related results canbe obtained using dynamic industrialevolution models with entry costs forformal sector participation, as I willargue in Section 4 below.

Descriptive country studies supportand elaborate upon this basic story.Writing on Peru in the 1980s, de Soto(1989) argues that many entrepreneursremained small to avoid excessive regu-lation, and thus failed to challenge en-trenched large firms, who moved rela-tively easily in the regulatory maze. OnCameroon in the early 1990s, Gauthierand Gersovitz (1997) show that smallfirms remained informal and avoidedtaxes, while large firms were influentialenough to obtain special treatment.Mid-sized firms bore the highest taxburden. On India, which is unusual inthe favoritism it has shown to smallfirms, Little, Mazumdar, and Page (1987,p. 32) write: “Not only would small firms[that graduate] have to cope with a muchmore difficult licensing policy, but theywould also have to contend with higherlabor costs (including wages and fringebenefits as laid down by labor laws) andsubstantially higher excise duties.”

These arguments help to explain thesize distribution in those developingcountries with heavy regulation. But thepervasiveness of small firms in LDCssuggests that other more universalforces are at work. For example, asnoted in Section 2, the poorest coun-tries tend to be the least urbanized, andtheir transportation networks tend to beunderdeveloped. So small, diffuse pock-ets of demand lead to small-scale, local-ized production.13 (This phenomenon is

central to some “big push” models of in-dustrialization, e.g., Kevin Murphy, An-drei Schliefer, and Robert Vishny1989.) In many countries a majority ofthe small-scale producers are located inrural areas, absorbing workers whenseasonal effects reduce agricultural em-ployment (Liedholm and Mead 1987,p. 28).

Underdevelopment also spawns smallfirms because Engel effects skew de-mand for manufactured products towardsimple items like baked goods, apparel,footwear, metal products, and furniture.All of these products can be efficientlyproduced using cottage technologies, sothere is little incentive to consolidateproduction in several large plants andincur the extra distribution costs.

Further, plentiful unskilled labor andthe lack of long-term finance create in-centives to economize on fixed capital.Since most machinery and equipmentmust be imported, the trade regime andthe lack of local technical support mayfurther militate against factory produc-tion in small markets.14 In the presenceof wage rigidities, abundant labor andscarce capital can also mean that formalsector jobs are rationed, hence workersunable to find employment in the for-mal sector may create their own micro-enterprises to survive.15

Finally, volatility in the business envi-ronment—both regulatory and macro-economic—can discourage mass produc-tion techniques. Investments in fixedcapital involve long-term commitmentsto particular products and production

13 Liedholm and Mead (1987) report that “in thefour survey countries where relevant data werecollected, direct sales to final consumers domi-nated [sales to businesses, government sales andexports], and, in fact, exceeded 80 percent inthree of the countries.” (pp. 46–47)

14 Cortes, Berry, and Ishaq (1987, pp. 153–54)note that “the increasing availability of skill ma-chine operators [in Colombia] has also contributedto the establishment of local importers and recon-structors of used equipment . . . ”

15 The evidence on this effect is mixed. Lied-holm and Mead (1995) argue that it is importantin sub-Saharan Africa, while Levenson and Ma-loney (1997) and Wendy Cunningham and Ma-loney (1998) downplay its significance in Mexico.


volumes. If there is substantial uncer-tainty about future demand conditionsfor these products, it often makes senseto choose production techniques that donot lock one in; that is, to rely moreheavily on labor (Val Lambson 1991;Brunetti, Kisunko, and Weder 1997).

3.2 Are Small Firms Scale Efficient?

Does the preponderance of smallfirms imply that scale inefficiency is aserious problem in developing coun-tries? Many have argued that it does,particularly in the simulation literature,where analysts often assume that the ra-tio of average to marginal cost is above1.10 for the typical plant.16 However,survey-based evidence suggests that thepotential efficiency gains from increasesin plant size—induced, for example, bytrade liberalization—are probably muchsmaller than these studies suggest.

The simplest studies of microenter-prises relate output per worker and out-put per unit capital to plant size. In ad-dition to the shortcomings of partialproductivity measures, this literaturesuffers from several data problems. Oneis that the boundaries of very smallfirms are often ill-defined because theyare part of a household or a farm, orbecause they are vertically integratedwith nonmanufacturing activities. An-other problem, stressed by Little(1987), is that many of these studiespool data on plants producing a diverserange of products. It is worth noting,however, that when Little, Mazumdar,and Page (1987) focus on four narrowly-

defined Indian industries, they find “itis difficult to detect any systematic vari-ation in labor or capital productivitywith firm size.” (p. 186)

Conceptually, studies of microenter-prises that attempt multi-factor produc-tivity measures are more appealing.One strand of this literature is based onsocial cost–benefit ratios, constructedas the cost of labor and capital atshadow prices, relative to value-addedin world prices. As discussed by Leid-holm and Mead (1987), these studieshave differed in their conclusions, withsome finding that small enterprises areat least as efficient as others, and othersfinding their efficiency relatively low.17

As for the very small, Liedholm andMead (1987) do find that one-personestablishments are systematically lessefficient than others, perhaps becausemany are created as occupations of lastresort for those who cannot find work inthe job market.

Scale economies are more consis-tently missing in studies of microenter-prises based on estimated productionfunctions. Little, Mazumdar, and Page(1987) and K. V. Ramaswamy (1994) fitsimple functions to cross-sectional dataon small-scale Indian producers, and re-port returns to scale very close to unityin all of the industries they treat. HalHill and K. P. Kalijaran (1993) obtainanalogous results among small-scaleIndonesian garment producers. Simi-larly, using firm-level African data col-lected by the Regional Program on En-terprise Development (RPED), Tyler

16 For example, Shanta Devarajan and DaniRodrik (1991) assume a ratio of 1.25 forCameroonian manufacturing; Drucilla Brown,Alan Deardorff, and Robert Stern (1991) assume aratio of 1.33 for most Mexican manufacturing in-dustries; and Jaime de Melo and David Roland-Holst (1991) assume ratios varying between 1.10and 1.20 for the Republic of Korea. Further de-tails are provided in Tybout and M. Daniel West-brook (1996).

17 Leidholm and Mead (1987) find that small en-terprises in Sierra Leone, Honduras, and Jamaicaare at least as efficient as others. On the otherhand, Sam P. S. Ho (1980) and Cortes, Berry, andIshaq (1987) find some evidence of scale econo-mies in Korea and Colombia, respectively. Smallenterprises are typically less capital intensive, soone reason for the discrepancy may be that Leid-holm and Mead use a rather high shadow price of20 percent per annum for capital services.


Biggs, Manju Shah, and Pradeep Sri-vasava (1995) fit the same estimator tofour manufacturing sectors in Ghana,Kenya, and Zimbabwe. Interestingly,even when the sample is limited tofirms with three to twenty workers, theyestimate returns to scale very close tounity. And when the entire stratifiedsample is used for each industry (cover-ing the entire size spectrum), returns toscale are still close to unity in food andtextiles/garments, while mild increasingreturns are found in wood products andmetal products.18

Finally, because data sets often donot cover the smallest plants, many ana-lysts have econometrically estimatedproduction functions using data onplants with at least ten workers. Theirdominant finding is constant or mildlyincreasing returns (between 1.05 and1.10) in the various manufacturing sec-tors of Latin American, Asian, andNorth African countries.19

All of the studies mentioned in thissection are plagued by measurementerror problems, omitted variables, ag-gregation bias, and simultaneity bias(Tybout and Westbrook 1996; JamesLevinsohn and Amil Petrin 1997; Tybout1992a). Nonetheless, their basic mes-sage seems consistent with engineeringstudies: the efficiency costs of being smallare not crippling—if present at all—once the one-worker threshold has been

traversed. Put differently, small firms indeveloping countries tend not to locatein those industries where they wouldbe at a substantial cost disadvantagerelative to larger incumbents.

4. Turnover, Market Share Reallocations, and Efficiency

Even if the potential gains from scaleeconomy exploitation are small, onemight argue that the prominence ofsmall-scale producers in LDCs is symp-tomatic of other problems. For exam-ple, if excessive taxation and regulationkeep many firms small and informal,these policies may be stanching theselection process through which bettermanagers and/or technologies gain mar-ket share.20 Severance laws and restric-tions on the use of temporary workersmay also inhibit the expansion and con-traction of formal sector plants, limitingcompetitive pressures. Similarly, pro-ducer turnover may be dampened bypolicies that prop up “sick” firms,thereby saturating the market with inef-ficient producers, and discouraging bet-ter firms from entering.21 Poorly func-tioning credit markets may furtherconstrain entry and expansion.

4.1 Analytical Models of Industrial Evolution

What might constitute evidence onthese relatively subtle effects? Dynamicmodels of industrial evolution provide18 This is all the more remarkable when one

considers that inherently inefficient firms tend tostay small, so even in the absence of scale econo-mies the data should exhibit some correlation be-tween size and productivity due to selection ef-fects (e.g., G. Steven Olley and Ariel Pakes 1996).

19 See Mark Pitt and Lung-Fei Lee (1981) onIndonesia; Brian Fikkert and Rana Hassan (1996)on India; Page (1984) on India; Tybout and West-brook (1995) on Mexico; Westbrook and Tybout(1993) on Chile; Tybout (1992a) on Chile; BrianAitken and Ann Harrison (1994) on Venezuela;Lee and William Tyler (1978) on Brazil; MonaHaddad and Harrison (1993) on Morocco; Tain-JyChen and De-PiaoTang (1987) on Taiwan; andBee-Yan Aw and Amy Hwang (1995) on Taiwan.

20 Of course, taxation and regulation are not in-efficient per se. As Levenson and Maloney (1997)note, firms that register with tax authorities andregulators also enjoy the benefits of enforceablecontracts, better access to credit, and—in the formof publicly administered fringe benefits for work-ers—access to risk-pooling mechanisms.

21 Pursell (1990) notes that “sick” enterprisespropped up by the Indian government tied uproughly 14 percent of total bank credit to industryin 1986. Fikkert and Hassan (1996) review thevarious licensing requirements and approval pro-cedures for capacity expansion that have prevailedin India, and provide further references.


some guidance. These models generallyinclude representations of the processesthat generate each firm’s entry, exit, pro-ductivity growth, and market share orfactor use. In most modern treatments,each dimension of performance is de-picted as the optimal behavior of forward-looking entrepreneurs with rationalexpectations but limited information.22

The literature is complex, but HugoHopenhayn (1992) provides a relativelytractable formulation. In his model,firms differ only in terms of their pro-ductivity levels, each of which evolvesaccording to an exogenous Markov pro-cess. New firms enter when the distri-bution from which they draw their ini-tial productivity level is sufficientlyfavorable that their expected futureprofit stream, net of fixed costs, willcover the sunk costs of entry. Firms exitwhen they experience a series of ad-verse productivity shocks, driving theirexpected future operating profits suffi-ciently low that exit is their least costlyoption. All firms are price takers, butthe prices of their inputs and outputsdepend upon the number of activefirms and their productivity levels.

This model shares a number of impli-cations with other representations of in-dustrial evolution developed by BoyanJovanovic (1982) and Richard Ericsonand Ariel Pakes (1995). At any point intime, an entire distribution of firmswith different sizes, ages, and produc-tivity levels coexists, and simultaneousentry and exit is the norm. Young firmshave not yet survived a shakedown pro-

cess, so they tend to be smaller and toexit more frequently. Large firms arethe most efficient, on average, so theirmark-ups are the largest. Nonetheless,despite all the heterogeneity, equilibriain both Jovanovic’s and Hopenhayn’smodels maximize the net discountedvalue of social surplus. Thus market in-terventions—like artificial entry barri-ers, severance laws, or policies thatprop up dying firms—generally makematters worse.23

Under certain regularity conditions,as Hopenhayn shows, an increase in thesunk costs of entry protects incumbentfirms from the upward pressure on in-put prices and the downward pressureon output prices that new entrants cre-ate. Thus high entry costs not only re-duce the amount of entry, they encour-age incumbents with relatively lowproductivity to stick around, andthereby increase the amount of produc-tivity dispersion among active firms.24

In addition, the market shares of thelargest, most efficient firms rise withentry costs, skewing the size distri-bution (Hopenhayn 1992, p. 1142). Theshares of the largest firms also respondnegatively to market size, since an out-ward shift in demand leaves the plantsize density function and the underlyingentry/exit processes unchanged.25

Policies that inhibit expansion or con-traction have similar consequences. Usinga variant on the model described above,Hopenhayn and Richard Rogerson(1993) simulate the effects of severance

22 Nelson and Sidney Winters (1982) argue thatmanagers do not have the knowledge or the timeto solve stochastic dynamic optimization problems,so these authors model entry, growth, and exit asderiving from rules of thumb that managers fol-low. The assumption of relentlessly optimal behav-ior is doubtless a caricature of the real world, butit is not obvious that alternative representations ofbehavior are more defensible. I will thus couch mydiscussion in terms of the optimizing literature.

23 Product markets are not perfectly competitivein Ericson and Pakes’s (1995) formulation so thisstatement does not hold for their model.

24 Exit costs have qualitatively similar effects tothose of sunk entry costs because they reduce theamount of one’s initial investment that can be re-covered by quitting the industry.

25 For example, if one doubles demand, concen-tration ratios drop by a factor of 2 and Herfindahlindices drop by a factor of 4, but turnover ratesand the market shares of each quantile in the sizedistribution remain unaffected.


laws. They find that increases in therate at which laid-off workers must becompensated raise the degree of persis-tence in firms’ market shares, increaseaverage firm size, increase price-costmark-ups, reduce average productivity,and reduce the job turnover rate.

If we think of sunk costs as derivingfrom formal sector entry rather than thecreation of a microenterprise, the re-sults I have reviewed above provide a

crude basis for inference.26 Specifically,they imply that among formal sectorproducers, low plant or job turnover,high mark-ups, and the frequent sur-vival of inefficient plants are symptomsof high sunk entry costs. Similarly,Hopenhayn and Rogerson’s (1993)simulations imply that persistence in

TABLE 2DETERMINISTIC FRONTIERS; AVERAGE EFFICIENCY LEVELS BY INDUSTRY

Industry

Little et al.(1987) or

Page(1984),India,

Translog,linear prog.

Cortes etal. (1987),Colombia,

Cobb-Douglas

Ramaswamy(1994),India,Cobb-

Douglas,linear prog.

Corbo andde Melo(1986),Chile,Cobb-

Douglas,linear

program-ming

Tyler(1979)Brazil,Cobb-

Douglas,linear

program-ming or

quadraticprogram-

ming

Pack(1984),

relative toUK bestpractice,

CES (subs.elasticity of

0.5)

Page(1980),Ghana,Cobb-

Douglaswith

gammadistribution

Food processing

0.58 0.401

Shoes 0.424 0.367Printing 0.645 0.334Soap 0.579 0.344Machine tools

0.688 0.56 0.432 0.372

Agricultural machinery

0.349 0.445

Plastic products

0.608 0.332 0.48 or 0.65

Steel 0.435 0.57 or 0.62Motor vehicles

0.638 0.312

Textiles: spinning

0.225 0.70(Kenya)

0.73(Philip-pines)

Textiles: weaving

0.68(Kenya)

0.55(Philip-pines)

Saw mills 0.370 0.710Furniture 0.743

26 Richard Caves (1998, pp. 1959–60) takes asimilar perspective on sunk costs in his discussionof entry patterns and hazard rates.


market shares and low turnover amongformal sector firms are symptoms ofbinding severance laws or restrictionson the use of temporary workers.

In addition to studies of the size dis-tribution of firms, at least three empiri-cal literatures provide evidence onthese symptoms. The first summarizesthe extent of productivity dispersion,usually in the context of efficiency fron-tier estimation. The second, relativelyrecent literature documents the extentof plant turnover, and in some cases re-

lates this turnover to productivitygrowth. Finally, an older literature onindustrial concentration is potentiallyrelevant. Let us take each in turn.

4.2 Is Productivity Dispersion Higher in LDCs?

Many analysts have studied theamount of productivity dispersion inLDCs. A sampling of results is pre-sented in Tables 2 and 3, and comparedto those from a recent multi-countrystudy of the OECD. Each study is done

TABLE 3STOCHASTIC FRONTIERS: AVERAGE TECHNICAL EFFICIENCY BY INDUSTRY∗

Ramaswamy(1994),

small- andmedium-

scale Indianfirms

Bhavani(1991), smallIndian firms

Hill andKalirajan,

(1993), smallIndonesian

firms

Biggs et al.(1995),Ghana,Kenya,

Zimbabwe

Pitt and Lee(1981),

Indonesianfirms∗∗

Tyler andLee (1979),small- andmedium-

scaleColombian

firms

Food 0.67 0.642Textiles and garments

0.626 0.46 (weavingonly)

min: 0.618max: 0.766

0.554(apparel)

Footwear 0.558Wood and furniture

0.42 0.984(furniture)

Metal products

0.719(structural

metal products)

0.51 0.987

Machine tools

0.727 0.704(agricult.

hand toolsonly)

Plastic products 0.820Motor vehicles 0.846

∗ All figures are estimates of E(e–u), where the inefficiency measure u is assumed to follow a half-normaldistribution.∗∗ Differences in estimated average efficiency reflect differences in the way that labor is measured, and whetherplant characteristics like size and foreign ownership dummies are included in the production function.


by estimating the “frontier” productiontechnology, which defines the maximumamount of output, y∗, attainable from agiven vector of inputs, x: y∗ = f(x). Then,for observed combinations of outputand inputs at the ith plant (yi, xi), theratio yi /f(xi) is interpreted either as anefficiency index itself, or as an effi-ciency index contaminated by measure-ment error and transitory shocks be-yond the control of plant managers.These two approaches are known as the“deterministic frontier” and the “sto-chastic frontier” approach, respec-

tively.27 Cross-plant average efficiencylevels and standard deviations in effi-ciency levels are the most commonlyreported summary measures of an in-dustry’s performance. These bear anegative monotonic relationship to oneanother in most cases, so I report onlythe former.

TABLE 3 (Cont.)

Caves et al.(1992)

Corbo and de Melo (1986),

Chilean firms

Kalirijan and Tse (1989),Malaysian

firms

Food processing 0.713 0.73Shoes wrong skewnessPrinting wrong skewnessSoap 0.627Machine tools wrong skewnessAgricultural machinery 0.751Plastic products wrong skewnessMotor vehicles wrong skewnessTextiles: spinning wrong skewnessApparel and weaving 0.649Saw mills 0.652Furniture wrong skewnessJapan, cross-industry average of 144 industries 0.699Korea, cross-industry average of 128 industries 0.672UK, cross-industry average of 72 industries 0.680Australia, cross-industry average of 91 industries 0.699US, cross-industry average of 67 4-digit industries 0.671

27 The literature can be further sub-divided ac-cording to whether f(xi) is estimated parametri-cally or non-parametrically (known as “data en-velopment analysis”), and whether econometric orprogramming techniques are used. WilliamGreene (1993) provides a recent summary of thevarious approaches to efficiency measurement.


Some caveats are in order. First,these studies are done at differing lev-els of aggregation. One would expectthat the finer the industry, the less dis-persion due to pooling heterogeneoustechnologies. Second, there are differ-ing degrees of measurement error inoutputs and inputs. Much of the cross-plant heterogeneity in capital and laboris unobserved by the econometrician,and most studies describe output interms of revenue rather than physicalproduct, blurring the distinction be-tween factor productivity and price-costmark-ups. Third, as is well known, theresults depend to a large degree uponwhether stochastic or deterministicfrontiers are used, and upon the as-sumed distribution of the error terms(Vittorio Corbo and Jaime de Melo1986).

Finally, unless they are estimatedwith panel data, stochastic frontiermodels separate technical inefficiencyfrom noise by treating ln[yi /f(xi)] as thesum of two orthogonal error compo-nents—one reflecting inefficiency andthe other reflecting measurement erroror shocks beyond the control of manag-ers. Typically, the negative of the ineffi-ciency component (hereafter denotedu) is assumed to have a half-normal,gamma or exponential distribution, andthe noise component is assumed to havea normal distribution. Greater skewness—measured by the negative of the thirdmoment of the compound error—thusimplies more productivity dispersion.However, the data often imply that thedistribution of ln[yi /f(xi)] is skewed in away that is inconsistent with these as-sumptions, so in practice many indus-tries do not fit the model. Such indus-tries are typically dropped from theanalysis, and the reported average effi-ciency levels are based only on theindustries that remain.

To control for differences in method-

ology, I have sorted studies according towhether they presume deterministic orstochastic frontiers, and wherever possi-ble, in the latter case I have used the re-sults based upon the half-normal distri-bution for the efficiency component ofthe error term. Among the deterministicfrontier studies, there is still some variationin the methodologies across studies be-cause some use linear programming toidentify the production function whileothers use quadratic programming. Moreimportantly, some (like Page 1980)impose a distribution on the efficiencymeasures, while others do not.

The deterministic frontier studies(Table 2) generally yield lower averageefficiency levels than the stochasticfrontier studies (Table 3), since the for-mer attribute all unexplained variationin y to inefficiency. Unfortunately, theyare also very sensitive to the specific as-sumptions behind the calculations, anddo not appear to convey any clear mes-sages. Notice, for example, that Corboand de Melo’s (1986) deterministicfrontier estimates imply that Chile wasvery inefficient relative to other coun-tries, but their stochastic frontier esti-mates imply Chile was average. Giventhis sensitivity, as well as the lack ofgood comparator studies from industri-alized countries, I shall hereafter focuson the stochastic frontier results.

Doing so reveals a surprising pattern.It is often observed that the cross-firmvariance in productivity levels is high indeveloping countries—e.g., Pack (1988),Evenson and Westphal (1995), MagnusBlomstrom and Ari Kokko (1997).Nonetheless, Table 3 suggests that aver-age deviations from the efficient fron-tier are not typically larger than whatwe observe in the high-income coun-tries studied by Caves et al. (1992). Thestandard methodology, when it “works,”yields mean technical efficiency levelsaround 60 to 70 percent of the best


practice frontier in both regions. Hencethe studies surveyed provide little sup-port for the view that LDC markets arerelatively tolerant of inefficient firms.28

One exception is provided by Biggs,Shah, and Srivastava (1995), who reportan unusually large amount of produc-tivity dispersion in Ghana, Zimbabwe,and Kenya. However, these results arebased on relatively broadly-defined in-dustries, so they may be a simple conse-quence of aggregation bias. In a par-ticularly detailed study, Pack (1984)finds average deviations among Kenyantextiles producers comparable to thosein other studies, even though his meth-odology is based on deterministic fron-tiers. Similarly, Page (1980) finds dis-persion levels typical of other countriesin his early study of Ghana.

Although the studies summarized inTable 3 do not find higher productivitydispersion in LDCs, they are not veryinformative. Most of them are based onoutdated methodologies. With a few ex-ceptions they rely on cross-sectionaldata, and hence must infer efficiencydispersion from the skewness of theproduction function residuals. Further,because most measure output as realrevenue, they misattribute cross-plantmark-up differences to productivity dis-persion. Finally, for lack of data, theytypically equate high productivity withsuperior performance, ignoring many ofthe costs that firms incur to enhancetheir technical efficiency.29 It is surelynot optimal to continually be the most

productive firm (George Stigler 1976).But without measuring the net presentvalue of firms’ efficiency-enhancing ex-penditures and the associated changesin their productivity trajectories, onecannot know whether observed patterns ofproductivity dispersion are problematic.

4.3 Plant and Job Turnover in LDCs

The literature on plant and job turn-over may be a better place to look forevidence on the strength of competitivepressures in the LDCs. If extensiveregulation and taxation combine withcredit market problems to keep smallfirms from challenging their entrenchedlarger competitors, we should observefew firms graduating from informal toformal status. Further, those firms thatdo graduate should show relatively littlemobility up the size distribution, andmarkets shares should be relativelystable among the largest firms.

Unfortunately, turnover figures areunavailable for many of the countrieswhere one would expect the policy re-gime to inhibit flux. However, a handfulof studies on Latin American, East Asianand North African countries provide somepreliminary evidence. These studies typi-cally document entry rates, exit rates,net job creation and net job destructionpatterns among the population of plantswith at least ten workers. (Variable defi-nitions are provided in the footnote toTable 4.) Most firms above the ten-worker threshold participate wholly orpartly in the formal sector, so measuredentry rates crudely describe formal sec-tor entry—both through new plant crea-tion and through the graduation of infor-mal plants to formal status.30 Similarly,the job turnover rates reported give us

28 Measurement problems make this finding allthe more remarkable. Noisy data–due to high andvariable inflation cum historic cost accounting—islikely to be more of a problem in LDCs, and thisshould exaggerate measured productivity disper-sion there.

29 These include training programs, employeerecruitment costs, technology purchases, andR&D expenditures. When labor inputs are mea-sured by work hours rather than by worker com-pensation, the wage premiums that firms pay toretain high-quality workers also belong on this list.

30 At that scale it is difficult to avoid detectionby the government, and the costs of forgoing busi-ness dealings with other formal firms and creditorsare substantial (e.g., Emilio Klein and Victor Tok-man 1996).


a crude sense of the stability of marketshares among formal sector firms.

Surprisingly, although Cox-Edwards(1993, p. ii) argues that Latin Americancountries “have a long tradition of try-ing to protect employment stability,”there appears to be more plant and jobturnover in these developing countriesthan others have found in the UnitedStates and Canada (Table 4).31 Over a

five-year interval, entering plants withat least ten workers captured 15 percentof the market in Chile, and entering

TABLE 4PLANT AND JOB TURNOVER IN DEVELOPING VERSUS DEVELOPED COUNTRIES∗

Turnover Rates Market Shares of Entrants

Country Plants Jobs <1 <5 <10 Minimum(period year year year Plant Sizecovered) 1 year 5 year 1 year 5 year olds olds olds Covered

Chile(1979–86)

8.5a — 26.9a — 3.6a 15.3a — 10 workers

Colombia(1977–89)

11.9a — 24.6a – 4.9a 19.8a — 10 workers

Morocco(1984–90)

9.5a — 30.7a — 3.2a — — 10 workers

Korea(1983–93)

— 64.2b — — — 32.5b — 5 workers

Taiwan(1981–91)

— 67.9c — — — 43.9c 63.2c 1 workerg

US(1963–82)

— 26.9d 18.9e 58.4f — 10.7d 18.6d 5 workers

Canada(1973–92}

— — 21.9e — — — — 5 workers

∗ Let Nt be the number of plants observed in year t; Et be the number of plants observed in year t but not t−1; andXt be the number of plants observed in year t−1 but not in year t. Then the entry rate is Et/Nt−1 and the exit rate isXt/Nt−1. The plant turnover rate is the average of these two statistics. Similarly, the rate of gross job creation is thenumber of jobs at entering plants plus the number of new jobs at expanding plants, divided by initial number ofjobs, and the gross job destruction rate is the number of jobs that disappear as plants contract or exit divided by theinitial number of jobs. The sum of these two rates is the job turnover rate.Sources:a Roberts and Tybout (1996), Tables 2.3, 9.3, 9.5, 10.3, 10.4, and 12.3.b Sukkyan Chung (1999), Tables 3.1 and 3.2.c Xiaomin Chen (1997), Tables 2.7 and 2.8; Aw, Chen, and Roberts (1997), Table 1.d Timothy Dunne, Roberts, and Larry Samuelson (1988), Tables 2, 3, 4, 8, and 9. Figures are average rates of newentry rates across 4-digit industries.e John Baldwin, Dunne and John Haltiwanger (1998), Table 1. These figures are for 1973–92.f Dunne, Roberts and Samuelson (1989), Table 1.g The Taiwanese data set describes firms rather than plants.

31 Among other distinctive features, Cox-Ed-wards (1993) notes that “The Latin American leg-islation, with a few exceptions, including Mexico,

is very strict in limiting the use of temporary con-tracts . . . firms cannot rely on a mix of perma-nent and temporary labor force, as is the case inJapan and increasingly the United States . . .”(p. 14). Hence it is difficult to avoid severancepayments by relying on temporary workers. On theother hand, as Cox-Edwards emphasizes, sever-ance payments are often legally tied to number ofyears on the job, so, subject to the temporaryworker constraints, firms may be encouraged to“maintain a very young work force with high rota-tion . . .” (p. iii)


plants with at least ten workers cap-tured 20 percent of the market in Co-lombia. On the other hand, in the moreinclusive population of plants with atleast five workers, entrants in theUnited States captured an average ofonly 10 percent of the market.32 Interms of job creation and job destruc-tion, Chile and Colombia average 27and 25 percent annual turnover rates,respectively, while the United Statesand Canada average 19 percent and 22percent, respectively.

Outside Latin America, some analystshave found even more flux in plants andjobs. In Morocco the annual manufac-turing job turnover rate was 31 percent.In Korea and Taiwan, over five-year in-tervals, new entrants captured an aver-age of 33 percent and 44 percent of themarket, respectively, compared to 10percent in the United States.33 Finally,although studies of the complete sizedistribution are unavailable for sub-Sa-haran Africa, Liedholm and Mead(1995) find that turnover rates amongmicro and small enterprises are veryhigh, ranging from 19 to 25 percent perannum (not in Table 4).

Why is there so much flux in theseLDCs? In some cases—especially inLatin America—high turnover partly re-flects the relatively dramatic businesscycles found there. In others—especiallyKorea and Taiwan—it partly reflectsrapid expansion of the manufacturingsector. But even if one focuses on the

minimum of the entry rate and the exitrate, turnover is relatively rapid in thedeveloping countries that have beenstudied.

Another part of the explanation lieswith Engel effects and low levels of hu-man capital, which encourage turnoverby skewing the output mix toward sim-ple products with relatively low start-upcosts, like baked foods, footwear, ap-parel, and metal products (Figure 2).The dominance of these sectors andtechnologies is probably amplified bymacro uncertainty, which creates incen-tives to be flexible in terms of productivecapacity.

Finally, policies seem to matter. Mar-ket share turnover rates are higher inKorea and Taiwan than they are inLatin America, where labor markets arerelatively regulated. Indeed, althoughambiguities remain, turnover rates ap-pear to be highest in Taiwan.34 Taiwan’slabor markets are least regulated amongthe sample countries (Joseph Lee andYoung-Bum Park 1995), and sunk entrycosts are relatively modest there be-cause the business environment makessubcontracting easy (Brian Levy 1990).

Of course, the turnover rates dis-cussed above don’t reveal which plants inthe population are expanding or contract-ing. It is possible that nearly all of theturbulence takes place among plants inthe 10–50 worker range, and that thesemoderately small producers never seri-ously challenge the larger, entrenchedincumbents. Indeed, one might arguethat high turnover rates in LDCs simplyreflect the relative importance of smalland medium enterprises there (Table1), and need not imply that large firms’market shares are more at risk.

32 Since turnover takes place mainly amongsmall plants, the contrast between these Northand South American countries would have beeneven greater if all studies had been done on acomparable basis. Turnover figures from Euro-pean countries are available, but they are reportedwith insufficient documentation to be useful forpresent purposes (John Cable and JoachimSchwalbach 1991).

33 The market share figures for Korea cover allplants with at least five workers, as are the UnitedStates figures. Note, however, that the Taiwanesedata describe all firms.

34 Comparisons of Taiwan with other countriesare unfortunately clouded by differences in sam-ple coverage (see previous footnote). Further evi-dence that Taiwan has relatively high turnover willbe introduced shortly.


For some of the LDCs that have beenstudied we cannot rule out this possibil-ity. For example, Chilean and Colom-bian plants with at least ten workers lost15 and 20 percent of their markets, re-spectively, over a five-year period (Ta-ble 4). But plants with less than 50workers account for more than half oftotal manufacturing employment inthese countries (Table 1), so all of themarket share loss could have come atthe expense of small producers.35 Thisis not true in Korea, where entrantscaptured 32.5 percent of the outputmarket and accounted for 46.5 percentof employment after five years, whileplants with less than fifty workers ac-counted for only 29.9 percent of manu-facturing employment. It is even lesstrue in Taiwan. There, plants observedin 1981 had lost 44 percent of the mar-ket to new entrants by 1986 and theyhad lost 63 percent of the market by1991, but plants with less than one hun-dred workers accounted for only 49percent of employment.

4.4 Turnover and Productivity Growth

High turnover does not necessarilyimply that inefficient producers are rap-idly driven from the market. For exam-ple, when the Argentine exchange rateregime collapsed in the early 1980s, itleft many firms with dollar-denominateddebt in serious trouble, and the result-ing exit patterns had little to do withproductive efficiency (Eric Swansonand Tybout 1988). It is therefore worthinquiring how well turnover “cleanses”LDCs of their least productive plants.

Several studies have quantified theeffects of exit and entry on sector-wideproductivity growth. Each is subject tothe measurement-error problems men-

tioned in Sections 3.2 and 4.2, but thefindings seem plausible and are gener-ally consistent with one another. InChile and Colombia, as in developedcountries, measured productivity amongexiting plants is much lower than it isamong incumbents (Lili Liu and Tybout1996; Liu 1993; and Tybout 1992b). In-deed, the productivity of exiting Chil-ean plants has often begun to deterio-rate several years before they actuallyexit (Liu 1993)—a phenomenon that ZviGriliches and Haim Regev (1995)dubbed the “shadow of death” effect intheir study of Israeli turnover. Simi-larly, Taiwanese plants doomed to dis-appear in the next five years exhibit be-low-average efficiency (Aw, XiaominChen, and Mark Roberts 1997). Sothere is evidence that a shakedown pro-cess is at work.

However, in Chile and Colombia, en-tering plants are also less productivethan incumbents on average. Further,neither entrants nor dying plants ac-count for more than 5 percent of totaloutput in a typical year.36 So inefficientplants are being replaced with plantsthat are only slightly more efficient, andneither group is a source of much pro-duction. This implies that if the turn-over process were suddenly arrested, theimpact on productivity would initiallybe small.

Nonetheless, over time the costs ofpolicies that prevent turnover quicklymount for several reasons. First, the“shadow of death” effect suggests thatexiting plants are on a downward trajec-tory, and might well continue to getworse. Second, entering cohorts typi-cally undergo a shakedown period in

35 For lack of information, I am ignoring the dis-tinction between employment shares and outputshares here. Among small plants the former typi-cally exceed the latter.

36 The low average productivity of enteringplants might seem at odds with the results I dis-cussed earlier which suggested small plants arenot much less efficient than large ones. These twofindings are not contradictory because, while mostnew plants are small, most small plants are notnew.


which the least efficient entrants dropout and the survivors quickly improvetheir productivity. Liu and Tybout(1996) find that this process brings theaverage productivity of new cohorts upto industry-wide norms after three orfour years in Colombia, and Aw, Chen,and Roberts (1997) find similar catch-up patterns in Taiwan, although theprocess is not complete there after fiveyears in some industries. Finally, whilethe firms turning over account for asmall share of production in any oneyear, the cumulative effects of turnoveron the population of plants quicklymount.

To empirically link the business envi-ronment with turnover-based produc-tivity gains, and to be rigorous about it,one would need to fit a dynamic struc-tural model of industrial evolution tofirm-level panel data. That has yet to beaccomplished, so in the meantime oneis tempted to look at cross-country cor-relations between productivity gainsand policies that affect turnover. Unfor-tunately, even this is problematic be-cause only two country studies have cal-culated turnover-based productivityfigures on a roughly comparable basis.We are thus left with the single tanta-lizing observation that turnover-basedproductivity growth has been higher inTaiwan (3.2 percent over five years)than in Colombia (2.2 percent over fiveyears).37 Those who are predisposed todo so might conclude that Taiwan’srelatively laissez-faire policies areresponsible.

4.5 Concentration, Price-Cost Margins, and Market Power

Industrial product markets are rela-tively concentrated in developing coun-

tries, and from this fact some observershave inferred—contrary to the turnoverstudies discussed above—that LDCslack competition.38 One rationale forthis inference is that high concentrationresults from high entry costs and insti-tutional constraints on labor markets,which limit the number of players andinsulate them from competitive pres-sures.39 Another rationale is that highconcentration in developing countriestraces to small domestic markets ratherthan restrictive policies, but it none-theless increases the sustainability ofcollusive arrangements.

Neither justification for inferringmarket power from concentrationseems compelling. The first rationaledoes not fit those developing countrieswhere evidence on turnover is available.High sunk entry costs and labor marketconstraints should depress turnover atthe same time that they increase con-centration, but we have seen that thesecountries show rapid flux in plants andjobs. The second rationale is correct in-sofar as market size is a powerful pre-dictor of concentration. Indeed, in LatinAmerica two-thirds of the cross-countryvariation in industrial concentration mea-sures is explained by the logarithm ofGDP alone.40 But concentrated markets

37 These figures are weighted averages of indus-try-specific results reported in Aw, Chen, andRoberts (1997) and Liu and Tybout (1996), re-spectively.

38 Norman Lee (1992) surveys the empirical lit-erature on concentration in LDCs so I will notrepeat the exercise here. Theorists have also beenknown to view high concentration in the LDCs assignaling relatively uncompetitive markets there(Paul Krugman 1989, Rodrik 1988).

39 Simulations of an industrial evolution modelwith noncompetitive market structures verify thatwhen countries differ because of sunk entry costs,differences in concentration are positively corre-lated with market power and monopoly rents(Pakes and Paul McGuire 1994).

40 This is the r2 I obtain using Patricio Mellor’s(1978) concentration measures, which were con-structed the same way for ten Latin Americancountries using their industrial census data. Anumber of studies have commented on the nega-tive correlation between market size and concen-tration—Lee’s (1992) survey provides further de-tails.


in the presence of vigorous turnoverneed not be conducive to collusivebehavior.41

If we eschew concentration as a sig-nal of market power, what other evi-dence is available? Some studies haveexploited Richard Schmalensee’s (1985)methodology, which “ . . . amounts toasking whether cross-plant variations[in price-cost margins] are due to indus-try-wide effects or to plant-specificmarket shares. Efficient plants shouldbe larger and have higher profits, so apositive correlation is generally expectedbetween market shares and price-costmargins, regardless of whether firmshave market power.” (Roberts and Ty-bout 1996, p. 196) On the other hand,persistent cross-industry variation inmark-ups, controlling for capital inten-sity, suggests that profit differentialsare not arbitraged away; that is, that en-try costs in some sectors limit competi-tion. There are many problems with thestudies that apply Schmalensee’s logicto LDCs, including the poor correspon-dence between their profit measures—price-cost margins—and economic prof-its (Franklin Fisher and John McGowan1983). But taken at face value, the re-sults for Chile, Colombia, and Moroccoshow no more evidence of market powerthan Schmalensee (1985) found in theUnited States (Roberts and Tybout1996).42

4.6 The Bottom Line

To summarize, because of institu-tional entry barriers, labor market regu-

lations, poorly functioning financialmarkets and limited domestic demand,the industrial sectors of developingcountries are often described as insu-lated, inefficient oligopolies. To date,however, there is little empirical sup-port for this characterization. Turnoveris substantial in the countries that havebeen studied, unexploited scale economiesare modest, and evidence of widespreadmonopoly rents is lacking.

The above notwithstanding, it wouldbe foolish to conclude that marketpower is a nonissue in developing coun-tries. Turnover studies and cross-plantstudies of profitability give one a gen-eral sense of the extent of competition,but they cover a limited and perhapsunrepresentative set of countries. Fur-ther, they are unlikely to detect isolatedpockets of noncompetitive behavior.For example, in Chile and Colombiaduring the 1970s, a handful of closelyheld conglomerates controlled largeshares of certain industries, as well asportions of the financial sector (Fer-nando Dahse 1979; Superintendencia deSociedades 1978). More recently, cozyrelationships between such conglomer-ates and the state have attracted atten-tion in East Asia (Ashoka Mody 1998).Finally, many developing countries haveprivatized natural monopolies duringthe past decade, and where efficientregulatory agencies have not sprung upto oversee them, noncompetitive prac-tices may be on the rise. Careful casestudies that collect detailed price dataand monitor the behavior of the individ-ual players are probably the only meansthrough which convincing conclusionsabout these problems can be reached.

5. Trade Protection, Market Structure,and Productivity

Even in those countries where compe-tition is vigorous, it is nonetheless im-perfect. Entry and exit costs matter, and

41 As noted earlier, Hopenhayn (1992) demon-strates the inverse relationship between marketsize and concentration in an industrial evolutionmodel with competitive product markets.

42 The country studies in Roberts and Tybout(1996) did find that the time series correlation be-tween margins and import penetration (trade bar-riers) was largest negative (positive) among bigfirms, suggesting that they are most directly incompetition with foreign suppliers.


products are differentiated. Further,learning spillovers and other externali-ties are surely present in some form.Hence protectionist trade policies,where they still exist, may do more thanaffect domestic relative prices and in-tersectoral resource allocations. Theymay change intra-industry mark-ups,productivity, or productivity growth. Inthis section, after briefly recounting therelevant theoretical literature, I con-sider the firm-level econometric evi-dence from LDCs on each possibleeffect.

5.1 Possible Effects of Trade Policy

Static arguments: There are numer-ous static arguments why trade protec-tion might affect the performance of do-mestic firms in LDCs. Most involve theeffects of trade policy on the competi-tive pressures that these firms face, thesize of the market that they operate in,or both. Firms’ responses often dependupon whether entry and exit barriers aresubstantial, whether scale economies—internal or external—are important, andwhether protection takes the form oftariffs or quantitative restrictions.43

I will limit myself to several exam-ples. Consider a tradeable goods indus-try with substantial entry barriers, com-posed of Cournot-competing firms. Ifthe industry has zero net exports underfree trade, the main effect of importprohibitions is to eliminate the threat offoreign competition. Domestic firmsmay exploit their enhanced marketpower by curtailing production and in-creasing their price-cost mark-ups, per-haps sacrificing some scale efficiency inthe process.

On the other hand, if the industry be-gins from substantial import penetra-

tion, the dominant effect of protectionmay be to increase the market size fordomestic producers. Firms are likely torespond by expanding, perhaps exploit-ing scale economies as they do so.(Mark-ups may still rise.) In either sce-nario, the higher profits that resultfrom protection may allow relatively in-efficient firms to survive, driving upproductivity dispersion. Alternatively, ifwe drop the assumption of prohibitiveentry barriers—which seems sensible,given our findings in Section 4—the higherprofits may eventually entice new, inef-ficiently small domestic producers toenter (Krugman 1979).44

External economies of scale furtherexpand the list of possible effects oftrade policy on productivity. Suppose,for example, that the external econo-mies occur at the industry level, and arenational rather than global.45 Then thenet effect of trade liberalization de-pends upon which sectors expand andwhich contract, as well as the magni-tude of traditional gains from compara-tive advantage effects. It is possible thatthe losses can outweigh the gains (e.g.,Helpman and Krugman 1985, ch. 3).

Finally, when employee effort is achoice variable, trade policy can affectthe amount of “managerial slack” or“X-inefficiency” among manufacturers.The dominant view among developmenteconomists is that protection inducesmanagers in import-competing industries

43 Many of the relevant models are summarizedin Elhanan Helpman and Krugman (1985) and anumber of the seminal contributions are collectedin Gene Grossman (1992).

44 Keith Head and John Reis (1999) summarizethe analytical and empirical literature on trade lib-eralization and the size distribution of firms, whileproviding some new evidence from Canada.

45 Industrial expansion generally deepens themarket for specialized labor, material inputs, andnetworked support services. Thus, even if no tech-nology spillovers take place, external scale econo-mies at the industry level may be present whenthere are increasing returns to scale in the produc-tion of these inputs, or when risk-averse special-ized workers prefer regions with many job oppor-tunities (e.g., Francisco Rivera-Batiz and LuisRivera-Batiz 1990; Francis Stewart and Ejaz Ghani1992).


to relax and enjoy the “quiet life.” Inearly versions of the argument, protec-tion increases profits among domesticfirms. This relaxes the consumption-leisure budget constraint faced by theirmanagers, who respond by choosingmore of both if they are on the back-ward-bending portion of their laborsupply schedule (W. Max Corden 1974).In more recent treatments, protectionaffects the payment schedule that own-ers (principals) must offer to managers(agents) to induce them to reveal theirendowed abilities.46 If the cost to own-ers of truthful revelation rises with pro-tection, they may opt for equilibria atlower output and effort levels, but theeffect of protection on effort is sensi-tive to modeling details (e.g., NeilVousden and Neil Campbell 1994).

Dynamic arguments: Further effectsof trade policy on performance havebeen demonstrated in explicitly dy-namic frameworks. Again, most any-thing can happen, depending uponmodeling assumptions and the particu-lar policy experiment. One issue thathas attracted attention is whether tradeprotection will induce technologicallybackward producers to invest in catch-ing up. In theory it may, if it increasesthe effective market size and the associ-ated pay-off from marginal cost reduc-tions for domestic firms (Kaz Miyagiwaand Yuka Ohno 1995; Rodrik 1992). Onthe other hand, protection may facili-tate collusion among domestic produc-ers and induce them to collectively stickwith backward technologies (Rodrik1992). A modest permanent quota mayalso delay technology adoption because,with continuously binding quantity con-

straints, foreign suppliers do not cutback their shipments to the domesticmarket when the home firm becomesmore efficient (Miyagiwa and Ohno 1995).

Catch-up models describe a one-timetransition from dated to new technolo-gies, but they do not link trade policiesto ongoing productivity growth. Forthat, theorists have developed generalequilibrium frameworks with continualknowledge production and diffusion. Insuch models, protection changes therelative prices of the inputs involved inproduct development, affects the set ofimported products that innovators com-pete with, and affects the ease withwhich domestic innovators can accessforeign technical expertise.

Whether protection reduces ongoingproductivity growth in these models de-pends partly upon the way in whichknowledge diffuses.47 Suppose thattrade policy does not affect the easewith which foreign knowledge can beaccessed, perhaps because it is readilyavailable over the internet. Further,suppose that to efficiently deploy tech-nical knowledge in LDCs, there is nosubstitute for learning-by-doing in thehigh-tech sectors and the spillovers itgenerates. Then trade protection mayimprove productivity growth and wel-fare if it promotes the high-tech activi-ties that generate the highest learningrates and the most valuable spillovers.48

On the other hand, if domestic pro-ducers acquire some of their knowledge

46 Given his or her endowed ability, each man-ager chooses an effort level in response to the re-ward structure and market conditions. By therevelation principle, contracts that induce manag-ers to be truthful about their (unobservable) abili-ties yield at least as high a value to the owner asany other mechanism.

47 Another key issue is the strength of spillovereffects. Diffusion of knowledge through technol-ogy purchases or licensing agreements is notenough to establish a link between steady stategrowth and trade policies. Knowledge spilloversare typically needed, and they must be strongenough that the private return to innovation doesnot fall with increases in the stock of knowledge(Charles Jones 1995).

48 Although they have not endorsed it, this argu-ment for protection has been formalized by Krug-man (1987), Nancy Stokey (1988), Alywn Young(1991) and Grossman and Helpman (1991).


through exposure to foreign clients,technologically sophisticated imports,or knowledgeable competitors, protec-tion may slow growth by constrictingimportant channels of knowledge trans-mission (e.g., Grossman and Helpman1991, ch. 6). Similar comments apply topolicies that discourage foreign directinvestment if the local presence of mul-tinational plants facilitates technologydiffusion.

Overall, the most striking conclusionthat emerges from the analytical litera-ture discussed above is that almost any-thing can happen when a country pro-tects its manufacturers, depending uponthe assumptions one invokes. Hencemany empiricists have attempted to de-termine what happens in practice bystudying patterns of association be-tween trade policy, pricing behavior,productivity, and productivity growth.Others have attempted to chip away atambiguities by asking which modelingassumptions best describe the data. Letus now consider the evidence.

5.2 Openness and Pricing Behavior: The Evidence

One of the more robust analytical re-sults on trade with imperfect competi-tion is that policies that constrain im-ports tend to increase the market powerof domestic producers. To look for evi-dence of this phenomenon, many re-searchers have regressed price-costmargins on proxies for import competi-tion or trade protection, usually lookingacross industries at a point in time.49

The correlation between import pene-tration (trade protection) and margins istypically negative (positive), and thestandard interpretation is that foreigncompetition squeezes monopoly rents,or “disciplines” the pricing behavior ofdomestic producers.

In a variant on this theme, a numberof authors have recently used RobertHall’s (1988) methodology for measur-ing mark-ups to gauge the effects of im-port competition on pricing behavior.50

The typical exercise is to regress outputgrowth on a share-weighted average ofinput growth rates, and interpret thecoefficient on input growth as a mono-tonic function of the price-cost mark-up. Allowing the coefficient to shiftwith trade liberalization, most studies inthis genre find that openness is associ-ated with reductions in price-cost mar-gins, and they interpret this to supportthe “import discipline” hypothesis.51

However, even if one ignores theeconometric problems, other interpre-tations are possible. Suppose that in-creased import-penetration reflects realexchange rate appreciation, whichsqueezes output prices relative to inputprices in the tradeable goods industries.In the short run, so long as revenuesstill cover variable costs, all firms—competitive or otherwise—will produceat lower margins and the competitivesectors will make negative economicprofits. Alternatively, suppose thatheightened import penetration reflectsthe removal of trade barriers. Thenrelative prices have been twisted in fa-vor of exportables and against import-ables. Stolper-Samuelson effects shoulddrive down the relative price of the in-put used intensively by the latter sec-tors, which is likely to be capital in thedeveloping countries. Similarly, incross-sectional studies, suppose that

49 Norman Lee (1992) surveys this literature.

50 Levinsohn (1993), Harrison (1994), FaezehForoutan (1996) and Pravin Krishna andDevashish Mitra (1997) study Turkey, Coted’Ivoire, Turkey (again), and India, respectively.

51 This methodology is likely to suffer from si-multaneity bias because transitory productivityshocks appear in the disturbance term, and arelikely to be correlated with input growth. Appro-priate instruments are nearly impossible to find(Thomas Abbott, Zvi Griliches, and Jerry Haus-man 1989).


protection is attracted to the industriesin the country’s comparative disadvan-tage, which are likely to be capital in-tensive. These same industries shouldexhibit relatively high price-cost mar-gins simply because relatively largeamounts of capital are used per unitoutput.

5.3 Openness and Productivity Levels: The Evidence

In addition to falling mark-ups, em-piricists tend to find that trade liberal-ization is associated with rising averageefficiency levels (Mieko Nishimizu andPage 1982; Tybout, de Melo, and Corbo1991; Tybout and Westbrook 1995; Har-rison 1996). Similarly, protected indus-tries tend to exhibit heightened produc-tivity dispersion (Tybout, de Melo, andCorbo 1991; Haddad 1993; Haddad andHarrison 1993). The standard interpre-tation of these results is that foreigncompetition drives inefficient domesticproducers to exploit scale economies,eliminate waste, adopt best practicetechnologies, or shut down.

However, as with the mark-up stud-ies, a number of caveats apply. First, si-multaneity bias creates the usual prob-lems. Inefficient, influential firms oftenlobby for protection, and sometimesthey succeed. Further, as already dis-cussed, in most firm-level data sets out-put is measured as revenue divided byan industry-wide deflator. So reductionsin measured “productivity” dispersionmay simply mean that mark-ups havefallen the most among firms with thelargest initial margins. That is, when ef-ficiency is equated with low dispersion—as it is in the efficiency frontier litera-ture—improvements in productivitycannot be distinguished from the mark-up squeeze often associated with tradeliberalization. Conversely, if trade liber-alization is associated with a major de-valuation, the favorable twist in prices

for tradeables should increase theirprofitability, at least in the short run.This looks just like an increase in aver-age efficiency among tradeable goodsproducers if physical units of outputcannot be observed.

Going beyond the association be-tween measured efficiency and open-ness, a number of authors have at-tempted to determine why the two arecorrelated. Their findings suggest thatinternal scale effects are not the mainreason. If trade liberalization forces in-efficiently small firms down their costcurves, one should observe plant sizesrising in import-competing sectors asprotection is removed. (This mechanismis built into the simulation models ref-erenced in footnote 16.) However, mi-cro panel studies consistently find thatincreases in import penetration are as-sociated with reductions in plant size, asare reductions in protection (MarkDutz 1996; Roberts and Tybout 1991;Tybout and Westbrook 1995). Thus lib-eralization may work against scale effi-ciency, at least in the short run. None-theless, the impact on efficiency ofthese plant size adjustments is probablysmall, since adjustments take placemainly among large plants that are op-erating in the constant returns range oftheir cost curves (Tybout and Westbrook1995).

Although external scale economiesare probably present, they too are un-likely to account for large protection-related efficiency effects. Using plant-level panel data and the methodologydeveloped by Ricardo Caballero andRichard Lyons (1990), Cornell J. Krizan(1997) finds significant external returnsto scale in many Moroccan industries.52

He then embeds his estimates in a

52 Other evidence that external scale effects arepresent comes from the locational choices of newfirms (e.g., Vernon Henderson and Ari Kuncoro1996).


computable general equilibrium modelof Morocco developed by ThomasRutherford, E. E. Rustrom, and DavidTarr (1993) and simulates the effects oftrade liberalization vis-à-vis Europe. Hefinds that external economies com-pound the gains from liberalizing, butthe effects are quite small.53 Further, toobtain large efficiency gains or lossesone would have to assume implausiblylarge external returns.

In sum, when trade liberalization im-proves productive efficiency, it is prob-ably largely due to intra-plant improve-ments that are unrelated to internal orexternal scale economies. The elimina-tion of waste, reductions in managerialslack, heightened incentives for techno-logical catch-up, and access to better in-termediate and capital goods are allpossible explanations, but there is littledirect evidence on the importance ofany of these. Detailed analysis of task-level efficiency and technologicalchoice within narrowly defined indus-tries—before and after a major changein trade policy—is probably the mostpromising direction for further work onthe topic.54

5.4 Openness and Productivity Growth: The Evidence

Static and dynamic effects of tradepolicy are conceptually distinct in thatthe latter involve a time dimension.However, all responses to policy taketime, even those that can be analyticallydescribed with a static model. Giventhe short time periods spanned by mi-cro data, it thus is rarely possible to dis-tinguish transitory one-shot adjustments

in productivity levels from lastingchanges in the rate of productivitygrowth. Hence, to assess the relevanceof the dynamic analytical models I men-tioned in section 5.1, I will limit my dis-cussion to the issue of how technologydiffuses.

Technology transfers through trade.As already noted, outward-orientedpolicies are more likely to facilitatelong-run growth if technology diffusesthrough international transactions. Forexample, LDCs may acquire new tech-nologies by de-engineering imports, orsimply by deploying the innovative in-termediate and capital goods that theyacquire in foreign markets. They mayalso learn about product design andnew technologies or management tech-niques from the foreign buyers to whomthey export. Once acquired throughthese channels, new foreign technolo-gies may diffuse to other domestic firmsnot directly engaged in trade. Are theseprocesses empirically important?

There is very little micro-econo-metric evidence on the productivity en-hancing effects of importing sophisti-cated intermediate and capital goods,although the fact that many LDCs im-port most of their machinery and equip-ment speaks for itself. Several studiesdo report a positive correlation betweenaccess to imported intermediate goodsand performance (Heba Handoussa,Nishimizu, and Page 1986; Tybout andWestbrook 1995), and Robert Feenstra,James Markusen, and William Zeile (1992)report evidence that Korean firms im-proved their productivity by diversify-ing their input bundles. Thus importedcapital and intermediate goods may bean important channel through whichtrade diffuses technology, but furtherwork is clearly needed to quantify theeffects.

More detailed evidence is available ontechnology acquisition through exporting.

53 In the scenario with the largest externality ef-fects, the positive effects of trade liberalization onwelfare increase from a 0.9 percent gain to a 1.1percent gain.

54 Pack (1984) and Mody et al. (1991) provideexcellent examples of research at this level of de-tail, but neither study directly examines the linkbetween performance and trade reforms.


In support of a “learning by exporting”effect, most studies that compare theproductivity of LDC exporters with thatof others in the same industry andcountry find that exporters do better.55

Case studies confirm that OECD buy-ers sometimes provide their LDC sup-pliers with blueprints and technicaladvice (e.g., Yung Rhee, Bruce Ross-Larson, and Garry Pursell 1984). Butfirms that are relatively efficient—per-haps for the reasons described in the in-dustrial evolution literature—are alsorelatively likely to self-select into for-eign markets (Sofronis Clerides, SaulLach, and Tybout 1998). Hence thecross-sectional correlation between ex-porting and efficiency may reflectcausality in either direction, or both.

Several recent studies attempt to ad-dress the causality issue by trackingfirms through time and asking, first,whether those that became exporterswere more efficient beforehand, andsecond, whether exporters showed im-provement relative to industry normsafter entering foreign markets. Mostfind that exporters were substantiallymore efficient than non-exporters be-fore they started selling abroad, so thehigher efficiency of exporters appears tobe at least partly a self-selection effect.56

These studies also find that in most in-dustries the efficiency gap between ex-porters and non-exporters does notgrow over time, suggesting that learning

is not a general phenomenon. However,firms in several industries do exhibitrelative efficiency gains after becomingexporters, so the learning-by-exportinghypothesis cannot be ruled out entirely.57

Regardless of whether they acquiretheir expertise from abroad, technologymay also diffuse from exporters to non-exporters in the same country, region orindustry through demonstration effects,skilled worker training (and subsequentlabor turnover), or expertise imparted totheir local suppliers. Looking at the in-tensity of exporting activity through time,Clerides, Lach, and Tybout (1998) findthat when many firms have been export-ing from a particular region, all firms inthat region tend to enjoy lower averagecosts. Spillovers are one interpretation,but this finding may simply reflect thefact that regions with cheap labor ormaterials are attractive export platforms.58

Technology transfer through FDI.Even if they are not innovative them-selves, multinational (MNC) affiliates inLDCs may transfer expertise to locallyheld firms through the same diffusionchannels I mentioned in connectionwith exporters. Indeed, FDI does seemto bring relatively efficient technologies

55 This result is reported in Aw and Hwang(1995); Aw and Geeta Batra (1998); Tain-Jy Chenand Tang (1987); Haddad (1993); Handoussa,Nishimizu, and Page (1986); Tybout and West-brook (1995); and Aw, Chen, and Roberts (1997).See, however, Rajeeva Sinha (1993).

56 Clerides, Lach, and Tybout (1998) report thisresult for Colombia, Mexico and Morocco (seealso their erratum at http://econ.la.psu.edu/~jtybout/clterr.pdf). Aw, Chen, and Roberts(1997) find the same pattern in Taiwan. Both setsof findings are consistent with those that AndrewBernard and J. Bradford Jensen (1999) report in arelated study of U.S. exporters.

57 Aart Kraay (1997) and Arne Bigsten et al.(1997) also find that firms become more efficientrelative to others after becoming exporters inChina and sub-Saharan Africa, respectively. How-ever, because of data limitations, their econo-metric models are relatively restrictive. In particu-lar, since the auto-regressive process generatingaverage costs is constrained to be first-order, theeffects of more distant cost lags may be comingthrough their lagged exporting dummies.(Clerides, Lach, and Tybout 1998 find that costprocesses are second or third order in Moroccoand Colombia.)

58 A different kind of productivity spillover fromexporters occurs if their activities ease the way forother firms to break into foreign markets. Demon-stration effects and the development of specializedsupport services like port facilities and intermedi-aries are possible reasons this might occur. Aitken,Gordon Hanson, and Harrison (1997) andClerides, Lach, and Tybout (1998) report someevidence that this phenomenon is present in Mex-ico and Colombia, respectively.


into host countries: most studies findthat foreign-owned firms are more pro-ductive than their domestically ownedcompetitors (Haddad and Harrison1993; Sinha 1993). But it is unclear howextensively these technologies diffuseamong domestically owned firms. Onthe one hand, case studies suggest thatsubstantial diffusion occurs (Blomstromand Kokko 1997). Further, firms in sec-tors with relatively high MNC presencetend to be more productive in Uruguay,Mexico, Morocco, and Venezuela (Kokko,Ruben Tansini, and Mario Zejan 1997;Haddad and Harrison 1993; and Aitkenand Harrison forthcoming). On theother hand, when industry effects arecontrolled for with dummy variables,domestically-held Venezuelan firms ac-tually do worse as the MNC presence intheir industry increases (Aitken and Har-rison forthcoming). Hence cross-sectionalstudies may suffer from simultaneity biasbecause MNCs are attracted to profit-able sectors, and negative spillover ef-fects may occur in the short run becauseMNCs siphon off domestic demandand/or bid away high quality labor whenthey set up shop in the host country(Aitken and Harrison forthcoming).

Learning-by-doing and learning spill-overs. As already discussed, theory tellsus that protection may facilitate produc-tivity growth by promoting domesticproduction in the learning-intensivesectors. Is this argument for protectionempirically relevant? In developingcountries, technology acquisition oftenamounts to adapting existing methodsto local circumstances (Evenson andWestphal 1995). Hence, instead of fo-cusing narrowly on R&D or technologypurchases, the rate at which firms gen-erate knowledge may be better proxiedby the intensity with which they rely onengineers, technicians, and scientists—hereafter ETS employees. If protectionencourages learning and productivity

growth, one would thus expect that ithelps ETS-intensive firms, and that thesefirms exhibit rapid productivity growthand/or generate positive spillovers.

However, it is not obvious that pro-ductivity growth and learning spilloversare greater among import-competingmanufacturers than among nontrade-able goods producers or export-orientedproducers. Arguably, the best docu-mented case of spillovers in LDCs isthe Green Revolution in Indian agricul-ture. Further, within each industry, thefirms that export—and thus the firmsthat benefit from openness—tend to bemore skill-intensive than others (GeetaBatra and Hong Tan 1997; Ana Revengaand Claudio Montenegro 1997; Clerides,Lach, and Tybout 1998).59

The presumption that ETS-intensivefirms exhibit the most rapid efficiencygrowth is also tenuous. Firms with highETS intensity do tend to get more out-put per unit bundle of capital and labor(Page 1980, 1984; Little, Mazumdar,and Page 1987; Cortes, Berry, andIshaq 1987; Biggs, Shah, and Srivastava1995). But the fact that ETS workersare more productive need not signalrelatively rapid learning-by-doing,much less spillovers from one firm toanother. In fact, in Colombia and Mo-rocco, ETS-intensive firms do not ex-hibit higher productivity growth thanothers (Julie Hunt and Tybout 1997).

Finally, although common sense andcase studies tell us that learning-by-doing among domestic firms is important,the available evidence suggests that itcomplements, rather than substitutesfor, access to the international knowl-edge stock (Evenson and Westphal

59 This is true despite the fact that their mar-ginal production costs tend to be lower, so it ap-pears that highly efficient firms hire the mostskilled workers and, because they are efficient,they also stand to gain the most from participationin foreign markets.


1995; Rakesh Basant and Fikkert 1996).In sum, the case for fostering growth byprotecting learning industries seemsweak.

6. Summary

The manufacturing sectors of devel-oping countries have traditionally beenrelatively protected. They have alsobeen subject to heavy regulation, muchof which is biased in favor of large en-terprises. Accordingly, it is often arguedthat manufacturers in these countriesperform poorly in several respects: (1)markets tolerate inefficient firms, so cross-firm productivity dispersion is high; (2)small groups of entrenched oligopolistsexploit monopoly power in product mar-kets; and (3) many small firms are un-able or unwilling to grow, so importantscale economies go unexploited.

The proliferation of very small plantsand the large market shares of bigplants in LDCs are sometimes inter-preted to support this position. How-ever, these distinctive features maysimply trace to the general economicenvironment. Small geographically dif-fuse markets and a demand mix skewedtoward simple consumer goods leadnaturally to large numbers of smallplants and to high concentration ratios.Indeed, although the issue remainsopen, the existing empirical literaturedoes not support the notion that LDCmanufacturers are relatively stagnantand inefficient. Turnover rates in plantsand jobs are at least as high as thosefound in the OECD, and the amount ofcross-plant dispersion in measured pro-ductivity rates is not generally greater.Also, although small-scale production isrelatively common in LDCs, there do notappear to be major potential gains frombetter exploitation of scale economies.

In many countries, therefore, themain manufacturing sector problemsmay not be of the variety that keeps

firms small, inhibits entry and exit,and/or creates market power. Rather,uncertainty about policies and demandconditions, poor rule of law, and cor-ruption may be the priority areas for re-form. These are certainly the areas thatmanagers identify as most problematicin qualitative surveys (Brunetti, Kis-unko, and Weder 1997). Also, for thosecountries that have not already done so,the removal of barriers to trade is likelyto improve efficiency. Falling price-costs mark-ups and rising productivityhave accompanied trade liberalizationepisodes in many LDCs.

As for future research, progress on anumber of fronts would seem especiallyuseful. First, given the fundamental im-portance of efficiency and productivitygrowth, improvements in the way thatwe measure these concepts should be apriority. For lack of detailed price in-formation, most of the work thus far hasequated physical output with real reve-nue, thereby blurring the distinctionbetween technical efficiency and profit-ability. Comprehensive reckonings ofthe costs of productivity gains are alsoneeded. The present value of trainingcosts, worker recruiting and retentionexpenditures, technology purchases,and research programs are seldom tal-lied up and weighed against the presentvalue of the productivity gains they gen-erate. But only this kind of calculationreveals firms’ economic efficiency.

Second, given the central role thatendogenous growth models assign tospillovers, any improvement in our un-derstanding of their form and magni-tude should help us to chip away at themystery of growth. This will requirebetter productivity measurement, asdiscussed above, and it will probablymean tracking individual firms over pe-riods of time long enough to deal withimpact lags. To the extent that technol-ogy diffusion takes place through labor


turnover, data sets that merge house-holds’ responses with those of their em-ployers would also be useful. Progressin this arena is likely to be gradual andpainful.

Third, studies that link firms’ behav-ior to uncertainty in the policy regimeand the macro environment are scarce,given the importance that LDC entre-preneurs attach to these phenomena. Ina similar vein, models that link labormarket regulations and regime volatilityto entry, exit, and productivity growthwould address some major unansweredquestions. Rigorous empirical analysisof these topics is difficult because it in-volves solving forward-looking optimiza-tion problems in the presence of uncer-tainty and sunk costs, sometimes withstrategic interactions among firms andpotential firms. Nonetheless, recenttheoretical work has laid some of thegroundwork for empirical modeling (e.g.,Ericson and Pakes 1995; Avinash Dixitand Robert Pindyck 1994) and estima-tion techniques are improving. The nextdecade is likely to bring real progress.

Clearly, all of these suggested direc-tions for research require improve-ments in the quality of data. Bettermeasurement of inputs (including train-ing, R&D, and other nontraditional fac-tors), outputs, and prices are needed ifwe are to have much confidence in find-ings on plant-level productivity mea-sures, or simply to document the incen-tive structure firms face at the groundlevel. More attention to data compara-bility across countries would also bewelcome. Unfortunately, the returns todata collecting and cleaning are verysmall because these activities do notdemonstrate cleverness to the econom-ics profession in any obvious sense.(They are sometimes interpreted todemonstrate the opposite.) Thus data-base building is generally underfunded,and in cases where the investments

have been made, the results have some-times been jealously guarded rather thandisseminated for widespread analysis.

REFERENCES

Abbott, T. A.; Zvi Griliches, and Jerry Hausman.1989. “Short Run Movements in Productivity:Market Power versus Capacity Utilization,”Harvard U.

Aitken, Brian; Gordon Hanson, and Ann Harrison.1997. “ Spillovers, Foreign Investment and Ex-port Behavior,” J. Int. Econ., 43:1–2, pp. 103–32.

Aitken, Brian and Ann Harrison. Forthcoming.“Do Domestic Firms Benefit from Foreign Di-rect Investment? Evidence from Panel Data,”Amer. Econ. Rev.

Aw, Bee-Yan and Geeta Batra. 1998. “Technologi-cal Capability and Firm Efficiency in Taiwan(China),” World Bank Econ. Rev., 12:1, pp. 59–80.

Aw, Bee-Yan and Amy Hwang. 1995. “Productivityand the Export Market: A Firm-Level Analysis,”J. Devel. Econ., 47, pp. 313–32.

Aw, Bee-Yan; Xiaomin Chen, and Mark Roberts.1997. “Firm-level Evidence on ProductivityDifferentials, Turnover, and Exports in Taiwan-ese Manufacturing,” NBER Working Paper6235.

Baldwin, John; Timothy Dunne, and John Halti-wanger. 1998. “A Comparison of Job Creationand Job Destruction in Canada and the UnitedStates,” Rev. Econ. Statist., 80:3, pp. 347–56.

Banarji, Ranadev. 1978. “Average Size of Plants inManufacturing and Capital Intensity: A Cross-Country Analysis by Industry,” J. Devel. Econ.,5, pp. 155–66.

Basant, Rakesh and Brian Fikkert. 1996. “The Ef-fects of R&D, Foreign Technology Purchase,and Domestic and International Spillovers onProductivity in Indian Firms,” Rev. Econ. Stat-ist., 78:2, pp. 187–99.

Batra, Geeta and Hong Tan. 1997. “Technologyand Firm Size–Wage Differentials in Colombia,Mexico, and Taiwan (China),” World BankEcon. Rev., 11:1, pp. 59–83.

Bernard, Andrew and Brad Jensen. 1999. “Excep-tional Exporter Performance: Cause, Effects, orBoth?” J. Int. Econ., 47:1, pp. 1–26.

Bhavani, T. A. 1991. “Technical Efficiency in In-dian Modern Small Scale Sector: An Applica-tion of Frontier Production Function,” IndianEcon. Rev., 31, pp. 149–66.

Biggs, Tyler; Manju Shah, and Pradeep Srivastava.1995. “Technological Capabilities and Learningin African Enterprises,” Africa Tech. Dept. Se-ries, Paper 288, World Bank.

Biggs, Tyler and Pradeep Srivastava. 1996. “Struc-tural Aspects of Manufacturing in Sub-SaharanAfrica: Findings from a Seven-Country Enter-prise Survey,” World Bank Discussion Paper346.


Bigsten, Arne; Paul Collier, Stefan Dercon, Ber-nard Gauthier, Jan Gunning, Anders Isaksson,Abena Oduro, Remco Oostendorp, Cathy Pat-tillo, Mans Soderbom, Michel Sylvain, FrancisTeal, and Albert Zeufack. 1997. “Exports andFirm-Level Efficiency in the African Manufac-turing Sector,” School of Business, U. Montreal.

Blomstrom, Magnus and Ari Kokko. 1997. “HowForeign Investment Affects Host Countries,”World Bank PRD Working Paper 1745.

Brown, Drucilla; Alan Deardorff, and RobertStern. 1991. “A North American Free TradeAgreement: Analytical Issues and a Computa-tional Assessment,” Econ. Dept., Tufts U.

Bruch, Mathias and Ulrich Hiemenz. 1984. Small-and Medium-Scale Industries in the ASEANCountries: Agents or Victims of Economic De-velopment? Boulder, CO: Westview Press.

Brunetti, Aymo; Gregory Kisunko, and BeatriceWeder. 1997. “Institutional Obstacles to DoingBusiness: Region-by-Region Results from aWorldwide Survey of the Private Sector,” WorldBank PRD Working Paper 1759.

Caballero, Ricardo and Richard Lyons. 1990. “Inter-nal versus External Returns to Scale in EuropeanIndustry,” Europ. Econ. Rev., 34, pp. 805–30.

Cable, John and Joachim Schwalbach. 1991. “In-ternational Comparisons of Entry and Exit,” inEntry and Market Contestability: An Interna-tional Comparison. Paul Geroski and JoachimSchwalbach, eds. Oxford: Basil-Blackwell.

Caves, Richard. 1998. “Industrial Organizationand New Findings on the Turnover and Mobil-ity of Firms,” J. Econ. Lit., 36:4, pp. 1947–82.

Caves, Richard; Sheryl D. Bailey, et al. 1992. In-dustrial Efficiency in Six Nations. Cambridgeand London: MIT Press.

Chen, Tain-jy and De-piao Tang. 1987. “Compar-ing Technical Efficiency Between Import-Sub-stitution-Oriented and Export-Oriented For-eign Firms in a Developing Country,” J. Devel.Econ., 36, pp. 277–89.

Chen, Xiaomin. 1997. “Resource Reallocation,Productivity Dynamics and Industrial Evolu-tion: Evidence from the Taiwanese Manufactur-ing Sector, 1981–91,” Unpublished dissertation,Pennsylvania State U.

Chung, Sukkyan. 1999. “Plant Turnover and Pro-ductivity Dynamics in Production and ExportMarkets: South Korea, 1983–1993,” Unpub-lished dissertation, Pennsylvania State U.

Clerides, Sofronis; Saul Lach, and James Tybout.1998. “Is Learning-by-Exporting Important?Micro-Dynamic Evidence from Colombia, Mex-ico and Morocco,” Quart. J. Econ., 113:3, pp.903–47.

Corbo, Vittorio and Jaime de Melo. 1986. “Mea-suring Technical Efficiency: A Comparison ofAlternative Methodologies with Census Data,”in Measurement Issues and Behavior of Produc-tivity Variables. Ali Dogramaci, ed. Boston:Kluwer Nijhoff Publishing.

Corden, W. Max. 1974. Trade Policy and Eco-nomic Welfare. Oxford: Clarendon Press.

Cortes, Mariluz; Albert Berry, and Ashfaq Ishaq.1987. Success in Small and Medium-Scale En-terprises: The Evidence from Colombia. NY:Oxford U. Press.

Cox-Edwards, Alejandra. 1993. “Labor MarketLegislation in Latin America and the Carib-bean,” Latin America/Caribbean Tech. Dept.,Report 31, Region Studies Program, WorldBank.

Cunningham, Wendy and William Maloney. 1998.“Heterogeneity in Small Scale LDC Enter-prises: The Mexican Case,” processed paper,The World Bank.

Dahse, Fernando. 1979. Mapa de la Extrema Ri-queza: Los Grupos Economicos y el Proceso deConcentracion de Capitales. Santiago: EditorilalAconcagua.

de Melo, Jaime and David Roland-Holst. 1991.“Industrial Organization and Trade Liberaliza-tion: Evidence from Korea,” In Empirical Stud-ies of Commercial Policy. Robert Baldwin, ed.Chicago: U. Chicago Press for NBER.

de Soto, Hernando. 1989. The Other Path: The In-visible Revolution in the Third World. NY: Har-per-Row.

Devarajan, Shanta and Dani Rodrik. 1991. “Pro-competitive Effects of Trade Reforms: Resultsfrom a CGE Model of Cameroon,” Europ.Econ. Rev., 35, pp. 1157–84.

Dixit, Avinash and Robert Pindyck. 1994. Invest-ment Under Uncertainty. Princeton: PrincetonU. Press.

Dunne, Timothy; Mark Roberts, and LarrySamuelson. 1988. “Patterns of Firm Entry andExit in U.S. Manufacturing Industries,” RANDJ. Econ., 19:4, pp. 495–515.

———. 1989. “Plant Turnover and Gross Employ-ment Flows in the U.S. Sector,” J. Lab. Econ.,7:1, pp. 48–71.

Dutz, Mark. 1996. “Oligopolistic Firms’ Adjust-ment to Quota Liberalization: Theory and Evi-dence,” in Mark Roberts and James Tybout,eds., op cit.

Ericson, Richard and Ariel Pakes. 1995. “Markov-Perfect Industry Dynamics: A Framework forEmpirical Work,” Rev. Econ. Stud., 62, pp. 53–82.

Erzan, Refik; K. Kuwahar, Sarafino Marchese, andRenee Vossenaar. 1989. “The Profile of Protec-tion in Developing Countries,” UNCTAD Rev.,1:1, pp. 29–49.

Evenson, Robert and Larry Westphal. 1995.“Technological Change and Technology Strat-egy,” in Handbook of Development Economics,Vol. 3. T. N. Srinivasan and Jere Behrman, eds.Amsterdam: North-Holland.

Feenstra, Robert; James Markusen, and WilliamZeile. 1992. “Accounting for Growth with NewInputs: Theory and Evidence,” Amer. Econ.Rev. (Papers and Proceedings), 82, pp. 415–21.

Fikkert, Brian and Rana Hassan. 1996. “Returns toScale in a Highly Regulated Economy: Evi-dence from Indian Firms,” IRIS India WorkingPaper 12, U. Maryland.


Fisher, Franklin and John McGowan. 1983. “Onthe Misuse of Accounting Rates of Return toInfer Monopoly Profits,” Amer. Econ. Rev.,73:1, pp. 82–97.

Foroutan, Faezeh. 1996. “Turkey, 1976–85: For-eign Trade, Industrial Productivity, and Com-petition,” in Mark Roberts and James Tybout,eds., op cit.

Gauthier, Bernard and Mark Gersovitz. 1997.“Revenue Erosion Through Tax Exemption andEvasion in Poor Countries,” J. Public Econ.,64:3, pp. 404–24.

Greene, William. 1993. “Frontier ProductionFunctions,” Working Paper EC-93–20, SternSchool of Business, NYU.

Griliches, Zvi and Haim Regev. 1995. “Firm Pro-ductivity in Israeli Industry,” J. Econometrics,65, pp. 175–203.

Grossman, Gene, ed. 1992. Imperfect Competitionand International Trade. MIT Press Readings inEconomics. Cambridge and London: MITPress.

Grossman, Gene and Elhanan Helpman. 1991. In-novation and Growth in the World Economy.Cambridge, MA: MIT Press.

Haddad, Mona. 1993. “How Trade LiberalizationAffected Productivity in Morocco,” PRD Work-ing Paper 1096, World Bank.

Haddad, Mona and Ann Harrison. 1993. “AreThere Positive Spillovers from Direct ForeignInvestment? Evidence from Panel Data for Mo-rocco,” J. Devel. Econ., 42, pp. 51–74.

Hall, Robert E. 1988. “The Relation betweenPrice and Marginal Cost in U.S. Industry,” J.Polit. Econ., 96, pp. 921–47.

Handoussa, Heba; Mieko Nishimizu, and JohnPage. 1986. “Productivity Change in EgyptianPublic Sector Industries after the ‘Opening,’1973–1979,” J. Devel. Econ., 20:1, pp. 53–73.

Harrison, Ann. 1994. “Productivity, ImperfectCompetition, and Trade Reform: Theory andEvidence,” J. Int. Econ., 36, pp. 53–73.

———. 1996. “Determinants and Effects of For-eign Direct Investment in Cote d’Ivoire, Mo-rocco, and Venezuela,” in Mark Roberts andJames Tybout, eds., op cit.

Hausmann, Ricardo and Micheal Gavin. 1996. “Se-curing Stability and Growth in a Shock-ProneRegion: The Policy Challenge for Latin Amer-ica,” in Securing Stability and Growth in LatinAmerica: Policy Issues and Prospects for Shock-Prone Economies. Hausmann and HelmutReisen, eds. Paris: OECD.

Head, Keith, and John Ries. 1999. “RationalizationEffects of Tariff Reductions,” J. Int. Econ.,42:7, pp. 295–320.

Helpman, Elhanan and Paul Krugman. 1985. Mar-ket Structure and Foreign Trade: Increasing Re-turns, Imperfect Competition and the Interna-tional Economy. Cambridge, MA: MIT Press.

Henderson, Vernon and Ari Kuncoro. 1996. “In-dustrial Centralization in Indonesia,” WorldBank Econ. Rev., 10:3, pp. 513–40.

Hill, Hal and K.P. Kalirajan. 1993. “Small Enter-

prise and Firm-Level Technical Efficiency inthe Indonesian Garment Industry,” Appl.Econ., 25, pp. 1137–44.

Ho, Sam P. S. 1980. Small Scale Enterprises inKorea and Taiwan. World Bank Staff WorkingPaper 384.

Hopenhayn, Hugo. 1992. “Entry, Exit and FirmDynamics in Long-Run Equilibrium,” Econo-metrica, 60, pp. 1127–50.

Hopenhayn, Hugo and Richard Rogerson. 1993.“Job Turnover and Policy Evaluation: A GeneralEquilibrium Analysis,” J. Polit. Econ., 101:5, pp.915–38.

Hunt, Julie and James Tybout. 1997. “Does Pro-moting High Tech Products Spur Develop-ment?” Processed paper, Georgetown U.

INEGI. 1995. XIV Censo Industrial. Aguas Cali-entes, Mexico: Instituto Nacional de Estadis-tica, Geografia e Infomatica.

Jones, Charles. 1995. “R&D-Based Models of Eco-nomic Growth,” J. Polit. Econ., 103, pp. 759–84.

Jovanovic, Boyan. 1982. “Selection and the Evolu-tion of Industry,” Econometrica, 50, pp. 649–70.

Kalirajan, K.P. and Y.K. Tse. 1989. “Technical Ef-ficiency Measures for the Malaysian FoodManufacturing Industry,” Developing Econ., 27,pp. 174–84.

Keller, Wolfgang. 1996. “Absorptive Capacity: Onthe Creation and Acquisition of Technology inDevelopment,” J. Devel. Econ., 49:1, pp. 199–227.

Klein, Emilio and Victor Tokman. 1996. Regula-tion and the Informal Economy. Boulder: Rien-ner Publishers.

Kokko, Ari; Ruben Tansini, and Mario Zejan.1997. “Trade Regimes and Spillover Effects ofFDI: Evidence From Uruguay,” StockholmSchool Econ.

Kraay, Aart. 1997. “Exports and Economic Perfor-mance: Evidence from a Panel of Chinese En-terprises,” World Bank Development Econ.Dept.

Krishna, Pravin and Devashish Mitra. 1997.“Trade Liberalization, Market Discipline, andProductivity Growth: New Evidence from In-dia,” Brown U.

Krizan, Cornell J. 1997. “External Economies ofScale in Chile, Mexico, and Morocco: Evidencefrom Plant-Level Data,” Ph.D. dissertation,Georgetown U. Econ. Dept.

Krugman, Paul. 1979. “Increasing Returns, Mo-nopolistic Competition and InternationalTrade,” J. Int. Econ., 9, pp. 469–79.

———. 1987. “The Narrow Moving Band, theDutch Disease, and the Consequences of Mrs.Thatcher: Notes on Trade in the Presence ofScale Economies,” J. Devel. Econ., 27, pp. 41–55.

———. 1989. “New Trade Theory and the LessDeveloped Countries,” in Debt, Stabilizationand Development: Essays in Honor of CarlosDiaz-Alejandro. Guillermo Calvo, ed. NY:Basil-Blackwell.


Lambson, Val. 1991. “Industrial Evolution withSunk Costs and Uncertain Market Conditions,”Int. J. Ind. Org., 9, pp. 171–96

Lee, Norman. 1992. “Market Structure and Tradein the Developing Countries,” in Trade Policy,Industrialization and Development: New Per-spectives. Gerald K. Helleiner, ed. Oxford:Clarendon Press.

Lee, Joseph and Young-bum Park. 1995. “Employ-ment, Labor Standards and Economic Develop-ment in Taiwan and Korea,” Rev. Labor Econ.Ind. Relat., special, pp. S223–S244.

Lee, Lung-fei and William Tyler. 1978. “The Sto-chastic Frontier Production Function and Aver-age Efficiency,” J. Econometrics, 7:3, pp. 385–89.

Leidholm, Carl and Donald Mead. 1987. Small-Scale Industries in Developing Countries: Em-pirical Evidence and Policy Implications. Inter-national Development Paper 9, AgriculturalEcon. Dept., Michigan State U.

———. 1995. GEMINI Action Research ProgramI, Final Report: The Dynamic Role of Micro andSmall Enterprises in the Development Process.Bethesda, Maryland: GEMINI.

Levenson, Alec and William Maloney. 1997. “TheInformal Sector, Firm Dynamics and Institu-tional Participation,” Processed, U. Illinois.

Levine, Ross. 1997. “Financial Development andEconomic Growth: Views and Agenda,” J. Econ.Lit., 35:2, pp. 688–726.

Levinsohn, James. 1993. “Testing the Imports-as-Market-Discipline Hypothesis,” J. Int. Econ.,35, pp. 1–22.

Levinsohn, James and Amil Petrin. 1997. “Produc-tivity, Industry Dynamics, and InternationalCompetition: A Structural Econometric Ap-proach,” U. Michigan Econ. Dept.

Levy, Brian. 1990. “Transactions Costs, the Size ofFirms and Industrial Policy: Lessons from aComparative Case Study of the Footwear In-dustry in Korea and Taiwan,” J. Devel. Econ.,34:1–2, pp. 151–78.

Little, Ian. 1987. “Small Manufacturing Enter-prises in Developing Countries,” World BankEcon. Rev., 1:2, pp. 203–36.

Little, Ian; Dipak Mazumdar, and John M. Page,Jr. 1987. Small Manufacturing Enterprises: AComparative Analysis of India and OtherEconomies. NY: Oxford U. Press.

Liu, Lili. 1993. “Entry–Exit, Learning and Produc-tivity Change: Evidence from Chile,” J. Devel.Econ., 42, pp. 217–42

Liu, Lili and James Tybout. 1996. “ProductivityGrowth in Colombia and Chile: Panel-basedEvidence on the Role of Entry, Exit and Learn-ing,” in Mark Roberts and Tybout, eds., op cit.

Lucas, Robert E. 1978. “On the Size Distributionof Business Firms,” Bell J. Econ., 9, pp. 508–23.

Maloney, William. 1997. “Labor Market Structurein LDCs: Time Series Evidence on CompetingViews,” U. Illinois.

Meller, Patricio. 1978. “The Patterns of IndustrialConcentration in Latin America,” J. Ind. Econ.,27, pp. 41–47.

Miyagiwa, Kaz and Yuka Ohno. 1995. “Closing theTechnology Gap under Protection,” Amer.Econ. Rev., 85:4, pp. 755–70.

Mody, Ashoka. 1998. “Industrial Policy after theEast Asian Crisis: From ‘Outward Orientation’to New Internal Capabilities?” processed paper,The World Bank.

Mody, Ashoka; Rajan Suri, Jerry Sanders, andDavid Van Zoest. 1991. “International Competi-tion in the Footwear Industry: Keeping Pacewith Technological Change,” World Bank In-dustry and Energy Dept., Industry Series WP51.

Murphy, Kevin; Andrei Schleifer, and RobertVishny. 1989. “Industrialization and the BigPush,” J. Polit. Econ., 97, pp. 1003–26.

Nelson, Richard and Edmund Phelps. 1966. “In-vestment in Humans, Technological Diffusion,and Economic Growth,” Amer. Econ. Rev.,56:2, pp. 69–75.

Nelson, Richard and Sidney Winters. 1982. AnEvolutionary Theory of Economic Change.Cambridge, MA: Harvard U. Press.

Ng, Francis. 1996. “A Profile of Tariffs, Para-Tar-iffs, Non-Tariff Measures, and EconomicGrowth in Developing Countries,” World Bank,International Trade Division, Int. Econ. Dept.

Nishimizu, Mieko and John Page, Jr. 1982. “TotalFactor Productivity Growth, TechnologicalProgress, and Technical Efficiency Change: Di-mensions of Productivity Change in Yugosla-via,” Econ. J., 92:368, pp. 920–36.

Olley, Steven and Ariel Pakes. 1996. “The Dynam-ics of Productivity in the TelecommunicationsEquipment Industry,” Econometrica, 64:6, pp.143–205.

Pack, Howard. 1984. Productivity, Technology,and Industrial Development: A Case Study inTextiles. NY: Oxford U. Press.

———. 1988. “Industrialization and Trade,” inHandbook of Development Economics, VolumeI. Hollis Chenery and T. N. Srinivasan, eds.Amsterdam: North-Holland.

Pack, Howard and Larry Westphal. 1986. “Indus-trial Strategy and Technological Change: The-ory versus Reality,” J. Devel. Econ., 22, pp. 87–128.

Page, John, Jr. 1980. “Technical Efficiency andEconomic Performance: Some Evidence fromGhana,” Oxford Econ. Pap., 32:2, pp. 319–39.

———. 1984. “Firm Size and Technical Effi-ciency: Applications of Production Frontiers toIndian Survey Data,” J. Devel. Econ., 16:1–2,pp. 129–52.

Pakes, Ariel and Paul McGuire. 1994. “ComputingMarkov-Perfect Nash Equilibria: NumericalImplications of a Dynamic Model Differenti-ated Product Model,” RAND J. Econ., 25:2, pp.555–89.

Pitt, Mark and Lung-fei Lee. 1981. “The Measure-ment and Sources of Technical Inefficiency inthe Indonesian Weaving Industry,” J. Devel.Econ., 9, pp. 43–64.

Pursell, Garry. 1990. “Industrial Sickness, Primary


and Secondary: The Effects of Exit Constraintson Industrial Performance,” World Bank Econ.Rev., 4, pp. 103–14.

Ramaswamy, K.V. 1994. “Technical Efficiency inModern Small-Scale Firms in Indian Industry:Applications of Stochastic Production Fron-tiers,” J. Quant. Econ., 10, pp. 309–24.

Rauch, James E. 1991. “Modeling the InformalSector Formally,” J. Devel. Econ., 35, pp. 33–47.

Revenga, Anna and Claudio Montenegro. 1997.“North American Integration and Factor PriceEqualization,” in Imports, Exports and theAmerican Worker. Susan Collins, ed. Washing-ton, DC: Brookings Institution Press.

Rhee, Yung; Bruce Ross-Larson, and Garry Pur-sell. 1984. Korea’s Competitive Edge: Managingthe Entry into World Markets. Baltimore: JohnsHopkins U. Press.

Rivera-Batiz, Francisco and Luis Rivera-Batiz.1990. “Effects of Direct Foreign Investmentin the Presence of Increasing Returns due toSpecialization,” J. Devel. Econ., 34, pp. 287–307.

Roberts, Mark and James Tybout. 1991. “Size Ra-tionalization and Trade Exposure in DevelopingCountries,” in Empirical Studies of CommercialPolicy. Robert Baldwin, ed. Chicago: U. Chi-cago Press for NBER.

Roberts, Mark and James Tybout, eds. 1996. In-dustrial Evolution in Developing Countries: Mi-cro Patterns of Turnover, Productivity andMarket Structure. NY: Oxford U. Press.

Rodrik, Dani. 1988. “Imperfect Competition,Scale Economies, and Trade Policy in Develop-ing Countries,” in Empirical Studies of Com-mercial Policy. Robert E. Baldwin, ed. Chicago:U. Chicago Press for NBER.

———. 1992. “Closing the Productivity Gap: DoesTrade Liberalization Really Help?” in TradePolicy, Industrialization and Development: NewPerspectives. Gerald Helleiner, ed. Oxford:Clarendon Press, pp. 155–75.

Rutherford, Thomas, E. E. Rustrom, and DavidTarr. 1993. “Morocco’s Free Trade Agreementwith the European Community: A Quantita-tive Assessment,” World Bank Working Paper1173.

Schiff, Maurice and Alberto Valdez. 1992. The Po-litical Economy of Agricultural Pricing Policy.Vol. 4: A Synthesis of the Economics in Devel-oping Countries. Baltimore: Johns Hopkins U.Press.

Schmalensee, Richard. 1985. “Do Markets DifferMuch?” Amer. Econ. Rev., 75, pp. 341–51.

Sinha, Rajeeva. 1993. “Foreign Participation andTechnical Efficiency in Indian Industry,” Appl.Econ., 25, pp. 583–88.

Steel, William. 1993. “Small Enterprises in Indo-nesia: Role, Growth, and Strategic Issues,” DSPWorking Paper 194. Jakarta Development Stud-ies Project II.

Stewart, Francis and Ejaz Ghani. 1992. “Externali-ties, Development and Trade,” in Trade Policy,

Industrialization and Development: New Per-spectives. Gerald Helleiner, ed. Oxford: Claren-don Press, pp. 122–54.

Stigler, George. 1976. “The Xistence of X-Effi-ciency,” Amer. Econ. Rev., 66:1, pp. 213–16.

Stokey, Nancy. 1988. “Learning by Doing and theIntroduction of New Goods,” J. Polit. Econ.,96:4, pp. 701–17.

Swanson, Eric and James Tybout. 1988. “Indus-trial Bankruptcy Determinants In Argentina,”with Eric Swanson, Studies in Banking and Fi-nance (supplement to J. Bank. Fin.) 7, pp. 1–27.

Superintendencia de Sociedades. 1978. Conglom-erados de Sociedades en Colombia. Bogota: Edi-torial Presensia.

Tybout, James. 1984. “Interest Controls andCredit Allocation in Developing Countries,” J.Money, Credit, Bank., 16:4, pp. 474–87.

Tybout, James. 1992a. “Making Noisy Data Sing:Estimating Production Technologies in Devel-oping Countries,” J. Econometrics, 53:1––3, pp.25–44.

Tybout, James. 1992b. “Linking Trade and Pro-ductivity: New Research Directions,” WorldBank Econ. Rev., 6:2, pp. 189–211.

Tybout, James; Jaime de Melo, and VittorioCorbo. 1991. “The Effects of Trade Reforms onScale and Technical Efficiency: New Evidencefrom Chile,” J. Int. Econ., 31:3–4–, pp. 231–50.

Tybout, James and M. Daniel Westbrook. 1995.“Trade Liberalization and Dimensions of Effi-ciency Change in Mexican Manufacturing In-dustries,” J. Int. Econ., 39:1–2, pp. 53–78.

Tybout, James and M. Daniel Westbrook. 1996.“Scale Economies as a Source of EfficiencyGains,” in Mark Roberts and Tybout, eds., opcit.

Tyler, William. 1979. “Technical Efficiency in Pro-duction in a Developing Country: An EmpiricalExamination of the Brazilian Plastics and SteelIndustries,” Oxford Econ. Pap., 31:3, pp. 477–95.

Tyler, William and Lung-fei Lee. 1979. “On Esti-mating Stochastic Frontier Production Func-tions and Average Efficiency: An EmpiricalAnalysis with Colombian Micro Data,” Rev.Econ. Statist., 61:3, pp. 436–38.

Vousden, Neil and Neil Campbell. 1994. “The Or-ganizational Cost of Protection,” J. Int. Econ.,37, pp. 219–38.

Wade, Robert. 1990. Governing the Market: Eco-nomic Theory and the Role of the Governmentin East Asian Industrialization. Princeton, NJ:Princeton U. Press.

Westbrook, M. Daniel and James Tybout. 1993.“Estimating Returns to Scale with Large, Im-perfect Panels: An Application to ChileanManufacturing Industries,” World Bank Econ.Rev., 7, pp. 85–112.

World Bank. 1993. The East Asian Miracle. NY:Oxford U. Press.

———. 1994. World Development Report 1994:Infrastructure for Development. NY: Oxford U.Press.


———. 1995a. Adjustment in Africa: Reforms,Results and the Road Ahead. NY: Oxford U.Press.

———. 1995b. World Development Report 1995:Workers in an Integrating World. NY: OxfordU. Press.

———. 1997. World Development Report 1997:The State in a Changing World. NY: Oxford U.Press.

Young, Alwyn. 1991. “Learning by Doing and theDynamic Effects of International Trade,”Quart. J. Econ., 106, pp. 369–405.


manufacturing firms in developing countries:how well do

Documents