volume title: annals of economic and social measurement, … · internac de distr. dcl ingreso,...
TRANSCRIPT
This PDF is a selection from an out-of-print volume from the National Bureau of Economic Research
Volume Title: Annals of Economic and Social Measurement, Volume 4, number 4
Volume Author/Editor: NBER
Volume Publisher: NBER
Volume URL: http://www.nber.org/books/aesm75-4
Publication Date: October 1975
Chapter Title: Production Functions for Industrial Establishments of Different Sizes: The Chilean Case
Chapter Author: Patricio Meller
Chapter URL: http://www.nber.org/chapters/c10420
Chapter pages in book: (p. 595 - 634)
.InnaIs / Ecu,mni (itu! .Soeja/ ,Ieasicri,iu'nt 44. 975
PRODUCTION FUNCTIONS FOR
INDUSTRIAL ESTABlISHMENTS OF DIFFERENT SIZES:THE CHftEAN CASE*
wr PATRIUJO MaLER
Lsing single eq uaiio'i ,nodel.s ICES and Cobb --Doughis), prodwi niH jiI1t( t(ijfls were t'',fini,ijed br ('/iih'anindustrial estahhshmenis : these estahli.thments were c/used according to their sizes. The purpose olthe paper is to examine if all esfahli.shnienss vi the same industr hare lilt' came or difierecii productionfunctions Chow tests and I/U' Iran.c!og production function were used to test this hi-pothesis .%hcrc'uit'r,the trend wit/i establishment SI:r of tile zec-hnohgicul parameters is examined.
I. INTROIuCTION
The main purpose of this paper is to study the technological characteristics ofindustrial establishments of different sizes. For each one of 21 Chilean industriesthe study analyzes the relationship between the size of establishments and thetechnology they utilize.
The main result can be summarized as follows: the empirical evidenceobtained supports the assumption of the existence of structural heterogeneitywithin the industrial sector--i.e., technology varies at the industry level as thesize of establishment increases. in other words, the industry is not a collection of"representative firms," and the size of establishment seems to be an importantelement in determining the technological characteristics of an industrial establish-ment.
The relation between establishment size and type of technology is examinedthrough the econometric estimation of production functions. The establishmentswithin each industry are first classified by size, and then production functionsare estimated for each size grouping of establishment. The breakdown of establish-ments by size groupings permits an examination of variations in technologicalparameters (product-productive factor elasticities, scale economies, elasticity ofsubstitution between factors) as establishment size varies. Such an examinationcan assist in answering several questions which have important economic policyimplications. How does the product-capital elasticity vary as the size of establish-ment increases? To what extent do economies of scale exist? What happens tothe elasticity of substitution between factors as the size of establishment increases(e.g., do isoquants become right angles or straight lines as we move farther awayfrom the origin)?
The validity of the methodology used, i.e., estimating production functionsfor establishments of different sizes within an industry, is verified by the Chowmethod and the translogarithmic production function method.
* This paper is a revised version of Chapter III of the author's unpublished Ph.D. Thesis "Pro-duction Functions and Efficiency Frontiers for Industrial Establishments of Different Sues. TheChilean Case. Year 1967," University of California. Berkeley. Jan. 1975.
595
The following paragraphs discuss the implicit theoretical framework employedto determine the technological characteristics ofdilferent firms within an industry.
Numerous economists have pointed out that a frequent phenomenon inunder-developed economies is the prolonged existence in a given economic sectorof firms employing substantially diverse production techniques. This phenoniençnis commonly called economic dualism.' Spaventa and others (see footnote I)have explained dualism in the following way. Some economies have had a non-homogeneous growth process in which an important segment of the system hasstagnated while the remainder has undergone sustained development. The conse-quence is the coexistence of two groups of firms with sharply different (l%'flamjccharacteristics which, in turn, have important implications for income distributionand labor absorption.2 The existence ofdiffercnt growth rates for different segmentsof an economy is, in itself, not a particularly striking fact. What is striking is theprolonged existence of dualism in underdeveloped countries.3 A wide range ofhypotheses have been formulated to explain the causes, consequences, and pro-longed survival of dualism.4 This study examines in some detail the phenomenoncalled intra-dualism, i.e., dualism within a given industrial sector as opposed todualism among different branches of industry or among diffesent economicsectors (e.g., agriculture versus industry).
The literature on economic dualism generally assumes an equivalenceamong size, modern technology, and efficiency. The prevalent premise is thatlarger establishments use more modern techniques and are more efficient thansmaller firms, but the equivalence among those three concepts is still an empiricallyopen question.
A. Pinto has generalized dualism by examining the feasibility and the conse-quences of more than two types of firms existing within an industrya situationknown as structural heterogeneity. Both Pinto and di Fillipo suggest that thecoexistence of different types of firms within a given industry results from theutilization of different technologies (where 18th and 19th century technologiescoexist with those of the 20th century). The technological differences amongfirms becomes a fundamental factor in explaining the labor productivity differ-entials among industrial establishments and the low rate of labor absorption by
C. Lutz, "The Growth Process in a Dual Economic System' (Quarterli Rtm'a. Sept. 958):L. Spaventa, "Dualism in Economic Growth" (Quarterlt' Rt't'iew. Dcc, 1959j S. H. Wellisi. TheCoexistence of Large and Small Firms: A Study of the Italian Mechanical Industries" (Quarti'rlvJournal of Eco,i.. Feb. 1957); T. Watanabe, "Economic Aspects of Dualism in the industrial Develop.inent of Japan' (Econ. Dc,. and Cult. Ch.. April 1965); A. Pinto. "Natuialeia c Implicaciones de a'heterogeneidad estructural' de Ia America Lalina" tTrirneszre Econ.. No. 145, Enero 1970): R R.Nelson, T. P. Schulti, and R. L. Slighton, Structural Change in a Det'elopnig Economy (Princeton UnivPress, 1971).
2 D. Turnham and I. Jaeger, The' Emp!orment Problem in Less Dcie'lope'd (4,i,nlrics (OECD.Paris l97lj; C.1.E.S.. "El Emplco y el Crecimiento en Ia Esirategia del Desarrollo de America Latinalmplicaciones para Ia Decada de los Setenia" (VII Reunion Anual dcl C.I.E.S.. Sept 1971)Leibenstejn, "Technical Progress, the Production Function and Dualism' (Quarte'rliRenew. Dec. 1960); T. Watanabe. op. cii.In addition to the bibliography of footnote I, additional bibliography with a survey on thesubject is found in H. Ellis, "Las Economias Duales y el Progreso" (Re'iista il" Eo,iomiu Ltino'americana No.3. 1961).
596
the industrial sector.S Structural heterogeneity as described by A. Pinto corres-ponds to R. Nelson's arguments that the use of a single production function tocharacterizerize an entire industry in an underdeveloped country conceals morethan it reveals, and that labor productivity differentials among industrial establish-ments cannot be simply explained by varying capital-labor ratios.ô
In operationalizing the structural heterogeneity hypothesis, the sue ofestablishment has been used as the fundamental variable in determining the typeof technology used by a fIrm. Furthermore, as previously mentioned, establish-ments of each industry are classified by size variable, and production functionsare estimated separately for each size grouping of establishment.
The study, finally, calls into question the utility of production functions as amicroecOnOrnic tool. l'he economic literature Contains many articles concerningthe variety of results obtained from econometric estimations of production func-tions. The sizable variations obtained in the values of the difl'erent technologicalelasticities (product-productive factors, economies of scale, factor substitution).should lead one to question their economic meaning. Why do the estimatedparameters have such great instability? Are there any economic hypotheses whichare not fulfilled'! If so, which are they?
The econometric estimation of production functions in this study permitscloser examination of the validity of economic policies based on the results ofproduction function analyses, particularly results concerning economies of scaleand elasticity of substitution.
Given the data available (i.e.. cross-section series), only a static analysis ispossible. Then when variations of different technological and economic variablesin the size groupings of establishment are examined, the consequences of thedualistic phenomenon at a given moment of time will be determined.
The data used in this study come from the Chilean Industrial ManufacturingCensus of 1967 disaggregated at the establishment level (11,468 establishmentsemploying 5 or more persons). For the purposes of this study, 21 industries willbe selected at ISIC International Standard Industrial Classification), four digitclassification level, and a separate analysis will be made for each of the 21 industries.
2. MErnouowGY
Different productive techniques are used by a set of firms of the same industryat a given point in time. 'rhe implicit assumption that firms in the same industryuse the same production function does not help to explain what happens inreality: on the contrary, it probably obscures a lot. Once the assumption ismade that all firms in the same industry use the same production function, theproductive techniques selected are a function of relative prices of the productive
A. Pinto y A. di Fillipo. "Notas sobre a Estralegia de Ia Distribuoo y Ia Redisribucion delIngreso en America Latina" (Semin. Internac de Distr. dcl Ingreso, Ceplan. mimec, Chile, Marzo1973).
6 R. R. Nelson et al.. op. cit.
597
factors, and the combination of these productive factors explains the (liflerencesin the productivity of labor.7
The working hypothesis utilized here is that the production functjoji variesfor establishments of different size.
The existence of different production functions within the same industrywill allow us to examine more closely some of the so-called "dual" characteristicsobserved in the industrial sector of an underdeveloped country. The term dualismis used here in the sense of intra-industrial, that is, to refer to the coexistence offirms using modern and old techniques in the same industry. The usual assumptionis that the largest firms are the most modern while the smallest are the mostbackward.8 With the existence of different production functions it is possible toexamine the empirical validity of the dualistic Concept.
The production function will be used as a tool to permit us to present in acompact form the most important technological characteristics of an industry.We will also try to identify the degree and types of variations of those technologicalcharacteristics for different size establishments within the same industry.
The method of estimating production functions used will be similar to thatused by Griliches and Ringstad,9 with some small variations. Each of the 21industries will be divided into 5 categories of different size-establishments: 5 to 9persons employed, 10 tc 19,20 to 49, 50 to 99, and 100 or more people employed,'0Then we will proceed to estimate production functions (Cobb-Douglas and CES)in each four-digit industry for each individual establishment grouping and for theindustry as a whole.
The criterion followed to estimate the different production function elasticitieshas been to employ a model which allows estimating each elasticity as a firstorder parameter. Even though every estimation model is different, all modelsused are very similar and it is assumed they are close approximations of the truemodel. Using that functional form which has a comparative advantage in theestimation of each elasticity is, in our judgment, the approach that will minimizethe unstable fluctuations of the estimators.
The Cobb-Douglas function was used to obtain estimates of the product-capital and product-labor elasticities, Kmenta's linearization of the CES was
1R. R. Nelson etal., op. ci:., pp. 91 92; B. S. Minhas. .4,, Inter,,,tjmu,j ('onrparLo,, ol hutor ("oiand Fu,'tor Use(North44olland 1963). pp. 30--3lTraditionally, the relationship between modern firm and large sue and backward lIrni andsmall size !s based on the following. New production techniques require a large solume of productionin order to take advantage of economies of scale they are very capital-intensive which means a largeinvestment, and are developed in the more industrialized countries in order to operate in larger markets.etc. Small firms use production techniques that are more labor-intensive, which is the factor moderntechnology tries to save, given the factor endowment prevalent in industrialized Countries which arethe ones in the vanguard of technological progress, J. Governeur. Productiri:v and Factor Proportionsin Less Dete!oped Counlrie.s(Clarendon Press, 1971); B. Singh. The &'oiumies of Small.S calt' !,,,1us,ri's(Asia Publishing House, 1961): NI. C. Shetty. Small-Scale and Household !sslusiries in a DecelopingEconomy (Asia Publishing House. 1963)Z. Griliches and V. Ringstad, Ecc,ioni it's of Scale and the Form f the Production Function(North-Holland, 1971).
'° See P. Meller, Ph.D. Thesis, op. ci!., a discussion of different variables that could be used forclassificatory establishment size purposes. The main empirical conclusion is that "when an establish-ment is large, ii is generally large in all its dimensions"_i,e. size groupings will not change very muchby the use of different size variables,
598
used for the economics of scale estimator. The traditional CES of Arrow et al.,''was used to obtain an estimator of the elasticity of substitution.'2
It is interesting to note that if either of the two traditional production functionsis used, i.e.. CES and/or CobbDouglas, the hypothesis that different-size firmspossess different production functions is compatible with the usual assumptionin economiC theory of the existence of a U-shaped average cost cur'e. The returnsto scale revealed by production functions of different-size establishment will bea test of the validity of the existence of a U-shaped average cost curve for anindustry.
The ordinary least-squares single-equation method is used to estimate theparameters of the production function. '"
The industrial establishments have been subjected to a rigorous selectionprocess. (See Appendix.) The criterion of selection has been the quality ratherthan the quantity of observations. The use of this criterion results in the exclusionof establishments in which some of the variables that are used in the econometricestimates are doubtful or whose data should be omitted.
In relation to the problem of measurement of variables after a series ofexperiments with alternative measurements, we have adopted a procedure similarto the one used by Griliches and Ringstad.'6 Labor requirements were measuredby the number of "equivalent" man-days employed by the establishment, andcapital requirements were measured by a flow of capital services obtained byusing book values (see footnote 16).
K. J. Arrow et al., "Capital-Labor Substitution and Economic Efficiency" (Re,. of Econ. andStat., August 1961).
12 Several other functional forms were estimated: Generalized CES form (J. Katz. ProductionFunctions. Foreign Inuestnient and Growth. North-Holland, 1969): I-lildebrand-Liu and Nerlove IM.Nerlove. "Recent Empirical Studies ol the CES and Related Production Functions." ed. M. Brown.The Theory and Empir. Anal, of Prod., N.B.E.R., l967) V. Mukerji("A Generalized SMAC Functionwith Constant Ratios of Elasticities of Substitution," Rev, of Ec. Stud.. 1963).
IS A. Walters, An Introduction to Econometrics (Macmillan, 1968), p.290.14 The estimators obtained by least-squares single-equation have many advantages: simplicity
of computation, small standard errors of the coefficients, and a high level of efficiency for prediction.But they also have many disadvantages, and the main one is that if the true model corresponds to asimultaneous equation model, then the least-squares single-equation estimators will be both biased and
not consistent. See A. A. Walters. "Production and Cost Functions: An Econometric Survey" (Econo-inetrica, Jan. 1963). pp. 18-22.
iS See P. Meiler, "Efficiency Frontiers for Industrial Establishments of Different Sizes" (Explora-lions in Economic Research, forthcoming), for an extensive discussion of the possibility ofestimating
production functions from cross-section data: i.e.. the factors explaining the existence of diffientproductive techniques in a particular industry are discussed.
16 Z. Grilches and V. Ringslad. op. cit., pp. 22-29. The variables are measured in the following
way:1: value added measuied in E-1967 (escudos of year 1967).L: labor factor measured in number of equivalent man-days, where the number of equivalent workers
is: L, = m1 (w2/w1)m2 + (2w2/w1)m3 in,. in2 and m3 are the number of blue collar, white collar
workers, and entrepreneurs. and w1 and w2 are the average wages received by blue and white
collar workers.K: capital factor measured in E°-l967, computed as a flow according to the following expression:
K = O.IOKM + 0.03K1 0.2OKr f 0.lO(Ku + K1 + K K,), where KM K1, K and K, are
the book values of machinery, buildings, vehicles and inventory goods. Linear depreciation rates
of 0.10, 0.03 and 0.20 have been used for machinery, buildings and vehicles, and a 10 percent real
interest rate is used as an alternative cost for immobilized capital.
599
LThe labor factor could have been measured through the number of man-days(designated by N1) without using that transformation to equivalent worker whichwould consider [he factor of difFerences in quality of labor. The simple co, relationcoefficients between L and N1 for the establishments within the 21 industrieshave values close to 1.0 (see Table I). This indicates it does not make any difFerencewhether the labor variable is measured by the number of man-days or by thenumber of equivalent worker-days.
TABLESIsu'I.E ('ORRELATION COFFIi('IENlS Bi rwi:i N nil ClrosN MFASURI:..MENTS ANI) AlTERNATIVE MEASUEEMFNTS OF Till LAIIoR ANt)
CAPITAL FAC1ORS SEPARA tEl) BY I NOUSI RY
Note L: number of equivaleni man-days: N1 number of man-days K flow of capital serices(see footnote 16); Ks,: book value ofniachinery KKW: number of KWh; K,,: number ofinslalled lip;K5: sum of the book values of machinery. buildings, vehicles and inventory goods
Different alternative measurements (proxy variables) could be used for thecapital variable. Among these are two flow variables: the previously definedcapital services, called K, and the number of KWh of consumed electricity, KK:also, the following stock variables could be used: the number of lIP installedcorresponding to the machinery related to the production process, K11, the totalbook value of the fixed assets plus the stocks of goods and inputs, measured inE° 1967, K, and lastly, KM, the book value of the machinery, measured in E°estimated to be the most reliable book value provided by the establishments.Above we present a table of simple correlation coefficients between K and thedifferent measurements of the capital factor for each one of the 21 industries600
Type ofIndustry
'SICCode L.N1 K:KM K;K K:K,, k:iç3111 0.969 0.959 0.806 0.597 0 9963112 0.992 0.983 0.892 0.874 0.9993116 0.981 0.972 0.622 0.680 0.9973117 0.984 0.970 0.866 0.837 09%3121 0.982 0.949 0.560 0.806 09973131 0.960 0.824 0.152 0.488 0.9993211 0.996 0.996 0.925 0.865 09983213 0.992 0.994 0.839 0.888 0.9953220 0.984 0.971 0.864 0.326 0.9973231 0.981 0.961 0.904 0887 09973240 0.989 1)985 0.575 0.641 09923311 0.988 0.913 0.675 0.707 0.9953320 0.985 0.942 0.875 0.875 0.9993420 0.884 0.992 0.949 0.672 0.9993560 0.990 0.994 0.927 0.880 0.9943693 0.984 0.936 0.558 0.926 09943710 0.989 0.989 0.697 0.840 09983813 0.989 0.986 0.887 0.536 0.9983819 0.983 0.989 0.810 0.867 09993829 0.976 0.992 0.728 0.790 09Q93843 0.973 0.832 0.864 0.389 0.995
(Table I). Most of the correlation coefficients are significant at the I percent level(see appendix for number of observations per industry).
Another variable needed for econometric estimation is w, average wages.In this case, w includes average wages received by blue and white collar workersworking in a given establishment, plus social legislation payments (employer'scontributions, bonuses, and child allowances) i.e., w represents the establishnient'scost of labor.
3. Esiistrtor'J OF 11Th VALUE ADDED-PR0I)ucTIvF FACTORS Ei ASTICITIES
To obtain estimators of elasticities of value added-producti.c factors onlythe CobbDouglas function was used because certain of its properties seem moresuitable in this case. Specifically, this means :1. ProductproductjvefactOrelas(icjtjescan be obtained directly as parameters of the first order: ii. Those elasticities areconstant for any production level (not the case for the CES) and iii. It is notnecessary to make assumptions regarding the type of market structure nor aboutthe firms decision-making policy.
We use the following notation for the Cobb-Douglas function:
Y = ALKwhere A, a, and are parameters of the functions and Y, L, and K corrcspondto aggregate value, labor, and capital respectively. Dividing this expression by Land denoting the parameter of economies of scale Iz a + fi I, we arrive atthe form traditionally used by Grilichcs :17
log = log A + b log + It log L.
The advantage of this form is that, in addition to obtaining the product-capital elasticity's directly, we also obtain directly the parameter It, i.e., theeconomies of scale. It is possible to verify directly whether the function has ordoes not have constant returns by testing the null hypothesis: if : h = 0.
The labor-capital elasticity, a, is found using the estimators of fi and !i,and the equation of definition, h = a ± fJ - 1. The estimator a found by thisindirect method is not biased if 1 and are unbiased estimators.'
The division of establishments into different size groupings was done todetermine whether variations occurred in different parameters of the productionfunction across different-size establishments. Let us see how fi varies as the sizeof the establishment is increased.
'Z. Griliches, "Production Functions in Manufacturing: Some Preliminary Results." TheTheory and Emp,rical Ana/v,js ofProducilon, op. cii., pp. 275-340; (iriliches and Ringstad, op. cit.,p.63.
IS Jf Y = .4LK0, Y/K = .4pL'K l
= (K/Y)(PYPK) = .4flLK'/Y ='
Let ft and 1 be the estimators for fi and I,. which are found by using the method ofteast squaresThen, if I + & - ft
E-E(l + -ft)= I +E1,-.Ejl= I + h ... =
601
(I)
(2)
The following table contains estimators for /3 calculated as a simple arithmeticaverage20 of the individual values of each of the 21 industries, for each sue groupof establishments
TAH[.l 2AVERA6I VAJUIS lop mr AooRii;ATr VAI.t:i:-CAi'iFAI 1tASJi(I I Y liv Si,'i 4)1
ES IAIIIISIIM NT
5(09 10 to 19 20 to 49 50(099 ll)orfllorepeople people people people people
As can be seen, the two largest size grouping of establishments have averagevalues far greater than the sma1er groups. This suggests that if the marginalproductivity of capital were t'ie same across firms, the smaller establishmentspossess a larger average Y/K (value added per unit of capital) than would thelarger establishments.2'It is difficult to find a definite trend of stable behavior for the large majorityof industries. However, in 16 of the 21 industries, the largest values fort? lie in thetwo classes containing the largest establishments, while in 15 ofthe2l industriesthe lowest values for /3 lie in the two classes with the smallest establishments.In the majority of the industries the largest establishments have larger valueadded-capital elasticity than the smallest establishments: however, there are caseswhere the situation is reversed. Even within the same industry the coefficient
varies considerably (see Table A-4): this behavior seems to suggest that the useof the same CobbDouglas function for a set of establishments within the sameindustry is not valid.Frequently, relative participation of the productive factors in the aggregatevalue, SK and SL are employed as estimators of the respective elasticities of outputwith respect to factor inputs, i.e. and This procedure is theoreticallyjustifIed assuming a firm has a CobbDouglas production function, enjoysconstant returns, faces competitive markets in goods as well as factors, and usesthe criterion of profit maximization. In this case:
= = /3 and '}L = =I will try different types of tests to check the hypothesis that the /3 estimatoris equal to the share of capital SK. Even though the different tests give contradictoryresults, my own conclusion is that there is no relationship between I? and SK.
20 have calculated the eIasticitj for each size grouping of establishments by using a simplearithmetic average of (he values obtained separately for each size group in each industry. Alternatemethods for obtaining this value would be (1) to find an estimator by using a regression for the set ofestablishments of a determined size group; (2) to calculate a weighted average of the values obtainedfor each industry separately. Either of these two alternative methods implies a different weight givento different industries. In this study I try to examine the characteristi of establishments of differentS1Zes in general and wish to avoid the influence of what might happen in one particular industry on thegeneral conclusions Thus, I gave the same weight to each industry, no matter what size it has.Negative values of f3were excluded, but their inclusion would not have changed the ranking of ft21 This would imply that if capital is the only scarce resource, small firms should be preferredbecause they maximize the output by unit of capital.
602
Average values of/ 0.401 0.34! 0.40k 0515 0
Table A-4 of the Appendix denotes values furnished by SK and fi. A secondtable where c is obtained indirectly was omitted because its values were veryunstable, even occasionally negative, while other values were much greater than1.0. Both of these situations are difficult to ilccept I[Oifl an economic point of
view.22In Table A-4 the values for S and fi are separated by industry for each one
of the five establishment size groupings and for the whole industry. It is difficultto determine an stable relation between values S and /3. In the following twoTables, 3 and 4. we calculated the coefficients of correlation between SK and j,separated by industry and by establishment size.
TABLE 3SistLL CORRELATION COEFFICIENTS BETWEEN
ANt) S SIPARA1FU IIY INDUSTRY
Type of Industry Correlation Coefficients
3111 0.600*
3112 -0.7903116 0.328*
3117 0704*3121 0353'3132 0.351'3211 0.291'3213 0.463'3220 0.071'3231 0.163'3240 0.184'3311 0.007'3320 0.340'3420 0.684'3560 0.475'3693 0.041*
3710 0.118'3813 0.677'3819 0.008'3829 0.091'3843 0.415'
* Correlation coefficient values not signi-ficant at 5 percent.
Finally, the correlation coefficient between (3 and SK taken at the industry
level for the set of 21 industries was 0.278.From Tables 3 and 4 and the previously given correlation coefficient. it is
indicated that there is no linear association between fi and SK. In other words. it
appears that (3 and SK measure different things, a result also encountered by
Griliches and Ringstad.23 However, in order to render a conclusive judgment.
I would test the null hypothesis: H0:fl SK for each one of the cases analyzed,
22 A negative value added-productive factor elasticity would stem from the fact that the marginalproductivity of that factor is negative. i.e.. a greater use of that factor implies a decrease in the value
added and would result in an irrational employment of that factor.Griliches and Ringstad, op. cit.. pp. 73-75. Griliches and Ringstad arrive at this conclusion
using rank correlation coefficients between a and S,.
603
J
5 to 9lOto 1920 to 4950 to 99100 and more
\Vhote lndustry
101Al.
tAitli 4SIMI'tt ('oRRtistt (IIIEII(II\I5 III Wi! N
AN!) S5 Sip.AitAlfI) BY ESIA!!! INJIMIN t Si,tSue of Establushnieni Correla to!) ( oellcjcn k
S to 9 person -- 0.055k10 to 19 persons OI)5220 to 49 peisons 0 299'SOlo99persons --0.121'tOO or more persons 0.295'
Correlation coefficient values not signilicantat 5 percent
NUMBER ot TIMES IIIATts A(cEt'TEI) OR RFJF(-j
Sue of Estabhshinent
TABLE 5nut: Nuti. FivIw titsis !f: /1
It) AT A Euvtj ot S!(;N:ti(AN(, o.5 PERCENu
/15 Rejected lmpos1ble to Reject !I
S
9644
IS1414
13
44 76
using the t values given in Table A-4.24 The results of this test are given in Table 5.This table suggests that it is not possible to reject the null hypothesis. Therelative share of the capital factor. S, is a good estimator of capital-aggregatevalue elasticity, at least for 63.3 percent of the cases. However, this hypothesis wasrejected in those industries or size groupings containing the largest number ofobservations Therefore, an increase in the number of observations in each casecould make possible the rejection of the null hypothesis H0 fl =r SK in a greatnumber of cases. From an empirical standpoint, the estimator of product-capitalelasticity S is inadequate for the industry as a whole, due to the rejection of thishypothesis in 12 0121 cases examined and at a significance level of 5 percent.is greater than estimator fi in 74 percent of the cases. Then, when SK is used asan estimator of product-capital elasticity, this elasticity is Overestimated in 74percent of the cases and underestimated in the remaining 26 percent of the cases.The magnitude of overestjnlatjon (and/or underestimation) fluctuates considerablyand no regular pattern of behavior emerged.But, even when the hypothesis H:f3 = S, is rejected for a large number ofcases, it does not imply the rejection of the basic assumptions such as perfectcompetition, constant returns, and profit maximization s, is empirically biased- Thj I test faces a serious objection due to the fact that S ts also a random ariahJe therefore
the i values obtained wilt be over-estimated making it easier to reject null hypothesis
604
I
upward because profits earned by the industry are iiiclttki When determiningthe relative shares of the capital factor in the aggregate tltii'. This is a partialexplanatiOr of why SK is greater than fin the maJorit of cases observed, I loweimperfections in the commodities and factors market also contribute a priorito make SK greater than II (in a short run model),25
4. EsriMvrio OF liii: E('oNo1fls 01: S'\1.E
The jaranieter for the economies of scale was estimated using tWO difl'ercntmethods: the Cobb-Douglas function and the linearization of Kmcnta of thegeneralized CES function.2
The notation for the generalized CIiS function is as follows
where ', . p. p are the CES function traditional paranletcrs.2Using the linearization of Knenta (expansion of(3) by a Taylor series around
the value of p = 0), the regression equation becomes:
log Y/L. = log ; + p log K/I. + It log L -F a0[iog KL]2
in which It p - I, being the scale elasticity. The expression (4) is equivalent to(2) plus the term a0[log K/LI2. Thus by testing the null hypothesis H0 a, - 0,we can determine whether the Cobb-Douglas function is or is not acceptable asan estimation model.
The economies of scale parameters estimated by the Cobb--Douglas andKrnenta functions are presented in Table A-5. In most cases the values obtainedwith the two indicators are highly similar. For this reason I need only examinethe type of variation evidenced by only one of them. I chose the Kmenta approx-imation because according to Monte Carlo's studies, the biases found in theestimator of the parameter for economies of scale arc insignificant when thisspecification is used.28
25 A model with imperfect competit ion in the market of goods and profit ma \Irnh/at ion h the
firm glcs: a/S, = qI = -- ,i sthere ; is the price eIasticit of the products demand. As and N, are
positive, for i to be negative t is necessary > N. We reach the same condition through a model of
imperfect competition in the labor market asailable to the firm. If we start from the ret9rns to scaleh and a and we calculated /1 (in this study the estimated elasticity), we woald obtain it 5 Fhis
explanation is not at all satisfactory from an empirical point of view, because 45. percent of the 't
elasticities would be "positise." (a has been indirectly obtained in this study IThe economies of scale parameter is obtained as a first order parameter. I his parameter is
constant throughout the whole estimation range and It S not necessary to make an h PC of issiiiiip-
tions about the market structure and or the firm's behavior.ix the efficiency parameter, ) is the distribution paranicter. p is he stihstituiioil parameter.
and p is the economies of scale parameter.1 G. S. Maddala and J. B. Kadana. "Fsfintion of Returns to Scale and the Elasticity of Sub-
stitution" (Econometrics., July 1967). pp. 421 422.
60
Y = '[( I -- O)L -I- oK
Excluding those values for economies of scale where IIj > 1.0. the figures inTable A-5 reveal the following information
TABLE 6NUMDFR OF EST:MA[ORS WHICH INI)i('ATi E(ONOSI!IS OR DISF(ONO,,1Ils (0 .Scr in
Sizr ot ESTARIISIMI Ni 1Ok i iw Sir OF 2! lNmsrkli s
Sue of t:siabi i.hnicniIn(luslr\
5 9 0 19 20 49 51) 99 1(X) Totalor more
Number of estimatorswhich mdicate economiesof scale
Number of estimatorswhich indicatedisccononiies of scale
2 3 5
16 5 IS S
11 we had estimated production functions for the industrial sector as awhole and for each industry separately, the conclusion inferred from the empiricalresults would be that nearly all Chilean industries are found to have economiesof scale. In some cases, these economies of scale reach 80 percent and their average(simp'e arithmetic mean) is 26.4 percent. These results might suggest that goern-mental policy should favor the appearance and survival of the largest establish-ments within the industrial sector in order to take advantage of these economiesof scale. However, this type of result is highly questionable, and below we willprovide some reasons to support our doubts. If we examine each industry separately,divided into different size groupings, and observe the Ti estimators obtained foreach one of those classes, the results are, in general, quite strange:
First, it is surprising to observe that even in the establishments of smallestsize, 16 estimators out of a total of 18 indicate the presence of diseconomies ofscale. A similar result was obtained in the two next larger classes: as a result, forthe three smallest size establishments (5 to 49 persons), a 10 percent increase inthe productive factors L and K, increases the aggregate value by less than 10 per-cent (for 15 of the 21 industries).
In a large number of industries each one of the size groupings evidencesa negative Ti estimator, although positive economies of scale are obtained for theindustry as a whole. One possible explanation for this phenomenon could be theone pictured in Figure 1, where we observe increasing average cost curves foreach size group of establishments and a decreasing average cost curve (economiesof scale) for the whole industry.
In various industries the Ti estimator increases with the size of establish-ment, but goes from negative values to positive values. This could be an excellenttest for proving that the average cost curve in various industries "takes the formof an inverted U."
An easy way out of all these strange results is to verify the null hypothesis:As shown in Table A-5, the null hypothesis H0:h = 0 is accepted in 77.8 percentof the cases considering only Ti estimators obtained by the approximation of
606
() 20
[
I
(5)
Figure I
Kmenta for different size groups. This suggests that the hypothesis of constantreturns to scale for each size class of establishments is rejected in only 22.2 percentof the cases (with a significance level of 5 percent). Meanwhile, at the level of thewhole industry, the hypothesis of constant returns to scale is relected in 14 of21 industries, with a significance level at 5 percent. Finally, as can be observedin Table A-5, in each size grouping almost 70 percent of the estimators indicatethe presence of diseconomies of scale. Diseconomies of scale prevail in Chileanindustry at the level of each size grouping of establishments.
In summary, not much confidence should be given to the results obtained forthe economies of scale parameter.
5. ESTIMATION OF THE ELASTICITY OF SUBSTITUTION
The traditional Arrow ci al. method (ACMS)29 was used to estimate theelasticity of substitution a.°
The estimation model is:
Ylog = a ± a log w.
The implicit economic assumptions in this model are that the establishmentoperates in a competitive market (goods and factors), maximizes profits, andthat there exist constant returns to scale.
Values of a will be provided at the total industry level and at the estabJishment-size level. At the end of this section, the trend that a has across an industry isdiscussed,
(i) Values ala at total industry level. For 14 o121 industries values of elasticityof substitution a fluctuate between (18 and 1.2. Except for 2 cases, the total of in-
Arrow et at., op. cii.° Two other methods were used to obtain estimators for o as pointed out previously the Katz
and Nerlove methods (see footnote 12). but they were rejected because of their theoretical limitationsand empirical inconsistencies. See P. MeIler's Ph.D. Thesis, o,. cit.. for a lull discussion of these issues.
607
V
\ (ILiStries has a values over 0.65. Only one negative value is obtained at the industrylevel and only 2 of the 20 positive values are not significant at 5 percent. (Table A-fl.)
(ii) Value.s oja at esiabjishpnent-s,ze let'el. Taking a simple arithmetic averageof the elasticity of substitution3' for each size grouping of establishment I obtainthe following results:
TABLE 7Anmisicric AVFRAGFS OF SuBsTITuTIoN ELASTICITIES FOR SI/f CiROtJi'IN(;S
OF ESTABLIShMENT
10059 10-19 20-49 50 99 and more
Values of a 0857 0839 0.764 0.670 I 266
There is a clearly decreasing tendency in the a values in establishments of5 to 99 persons (momentarily excluding the largest Ones). Considering these fivesize groupings, we can think of a U -shape relation between the a values and thesize groupings of establishments. This coincides with the results obtained by Abein the study of Japanese firms.32 Establishments of 50 to 99 persons probablyhave a more inflexible technology, while larger establishments probably use moreflexible techniques.
This empirical result also supports Leibenstein's thesis33 that for any giventype of technology, the instrument of relative prices will be more likely to affectthe choice of techniques in larger establishments than in smaller ones, becausesmall variations in the relative prices of productive factors will produce a relativdygreater impact in those establishments using more elastic techniques from thesubstitution point of view.
This conclusion requires the following qualification: the first four sizegroupings of establishments from 5 to 99 persons could be thought of as corres-ponding to size groupings of establishments approximately comparable to eachother in terms of range of technology. Meanwhile, the firms employing 100 ormore persons are an open grouping of fairlyextensive range, clearly different fromthe range covered by the other four. Therefore, it was possibl'e to predict a priorihigh values obtained for this large grouping.
(iii) Values of a in each industry. It is interesting to know the tendency theelasticity of substitution has in each particular industry. Six possible tendenciesof a related to the increase of establishment size groupings were chosen : increasing,decreasing, constant, (all values no different in more than 0.2), fl-form, U-formand indefinite. Those industries with less than 4 values of a were excluded (theyeither had negative values or had a small number of observations).
Negative values of a have been excluded, hut if they were considered, the ranking of a valueswould not change.- M. A. Abe, "The Growth Pati, of Firms and the Development Process of the Economy: TheCase of Japan" (The Developing Economies, June 1972), p.201.H. Leibensje,n "Technical Progress, the Production Function and Dualism" (Bunco Nu:ionh'del Lacoro Quarter!;: Review, March 1960), pp. 348-351.
608
TABLE 8Ntllnik OF INI)USLkIIs HAVINGA DF1IBM1NEI) TFNDIN(y Br-I WiEN Ci AM) Stir GROUPIN(;S
(JJ ESFAIII!SIIMrNT
Excluding largeTendencies estahlishrnem
of groupings
IncreasingDecreasingConstantfl-form 3U-form 4indefInite 6
There is no definite tendency for a to vary uniformly across the different sizegroupings. Furthermore, the elasticity of substitution does not remain constantas the size of establishment varies, suggesting that the isoquant map is not honio-thetic.
Considering the magnitudes taken by the elasticity of substitution, Table 9provides the number of a values obtained by size groupings and range of values.
TABLE 9NIJMnrR OF ELASTICITY OF SUIISTITUJION VAluEs sy SizrGouptNGs oi Esr.misntrns ANn RANUF OF VAlUES
The values in this table agree with the condensed and analyzed data ofTable 7.
(iv) Tests ofa for values 0 and I. The traditional tests made with the elasticityof substitution correspond to the hypothesis a = 0 (fixed proportions function)and a = I (CobbDouglas function). In the case H0 :a = 0, Table A-6 shows thatthis hypothesis is rejected in 83 of 120 cases (69.2 percent) at a significance levelof 5 percent. These results indicate that Chilean industry, at a level of differentestablishment size, does not present a rigid technological structure of fixedproportions. I could also calculate the value of the statistics t in Table A-6 forthe null hypothesis H0:a = I. However, taking advantage of the Kmenta ap-proximation, it is simpler to use a direct test where the quadratic term parameterpermits immediate verification of the null hypothesis by showing whether ornot the production function is Cobb-Douglas. This test for the different sizegroupings of establishments shows that the null hypothesis (the production func-tion is not CobbDouglas) is rejected in 7 out of 120 cases (58) percent) at a sig-
609
Sizegroupings
Range of Values
0.00-0.50 0510.80 0.81 -1.20 I-1.21 and more
5-9 3 5 8 310-13 2 8 5 320-49 6 7 4 350-99 8 2 5
l00andrnore I I 6 6
;------
(6)
nificance level of 5 percent. As a result. I do not have to discard the conclusions andmagnitudes previously obtained for the product-capital elasticity and thceconomies of scale, which critically depend on the condition that the Cobb-Douglas function be consistent with the data.
6. HoMorneTiciTy TESTS OF THE PRootrcTIoN FUNC-nON
The central point to be examined in this section will be to find out whetherthere is a structural change in the technological parameters of the productionfunction when I go from one size grouping of establishments to the next.
In the previous sections, where I examined the estimated values for theproduct-capital, economies of scale, and substitution elasticities, I affirmed theseparameters were subject to change for different size groupings. This alreadysuggests that a production function with constant elasticity is not an adequatetool for synthesizing the technological characteristics of the establishments ofdiffering sizes in the same industry.
The traditional econometric procedure for examining the hypothesis ofstructural change in the parameters is the Chow test. An alternative way toexamine the null hypothesis as to whether the production function is homotheticis using a heterothetic function, the trans-logarithm production function.34 Inthis study I will use these two procedures to verify the homothetiticity of theproduction function.
(a) Chow Test
To examine the hypothesis of a structural change in the parameters, a generalChow Test will be used. This test attempts to verify whether the breakdown ofthe establishments into 5 size groupings is significant or noti.e., the null hypo-thesis showing no difference between the five size groupings. If I assume thatvector /3. represents the vector of technological characteristics of size grouping 1,the H0 could be written: Ho:fl = /32 = /33 = [34 =
The alternative hypothesis is that the five size groupings are different fromeach other. This implies that for each size grouping of establishments thereprevails a different production function, or that the production function for theindustry does not possess constant elasticities (product-inputs, economies ofscale, and substitution).Let K be the number of explanatory variables for the regression and N thenumber of observations
Q1 will be the sum of the squared residuals of the regressionfor size grouping i, and QT the sum of the squared residuals for the whole industry(the five size groupings Simultaneously).The Chow Test for testing the hypothesis H0 takes the following expressionfor the case of five subsamples
QT-1Q N-5(K+ 1)FG = -= K + 1) -'
L. R. Christensen, D. W. Jorgenson, and L. J. Lie, "Transcendental Logarithmic ProduclionFunctions" (Rev, of Econ, and Stat, Feb. 1973).D. S. Huang, Regression and Econometric Methods (ViIey and Sons, 1970),
610
-
- .1- -.: -_._-
*
whereFGhaSadistribution Fwith4(K ± 1)and N - 5(K + l)degreesoffreedom.Table 10 provides the F4 values for the three different functional forms used inthis study.
TABLE 10F STAJISTICS OF CHOW TEST FOR EACH INI)USTRV AND
FOR IIREE DIFFERENT FUNCTONA1. FORMS
Values of F not significani at 5 percent.
Table 10 indicates that the null hypothesis according to which no differencesexist between technological parameters of establishments of different sizes isiejected in over 60 percent of the industries considered.
(b) The Trans-logarithmic Production Function
The trans-logarithmic production function is obtained by expanding thequadratic term of the linear approximation of Kmenta:
log = a0 4 a, log + a2 log L ± a3(log K - log L)2
Y Ka0 + a1 log + a2IogL + a31(IogK)2 + a32(logL)2
(7) -2a33(log K)(log L)
The trans-logarithmic function is heterothetic, while the Kmenta approxima-tion is homothetic. To verify the homothetic hypothesis for the productionfunction, Griliches and Ringstad suggested three alternative forms,36 one of
Grilichcs and Ringstad, op. ci:., pp. 10 and 88.
611
Type olIndustry
ACMSFunction
Cobb Douglas Kmenta'sFunction Linearization
311 I 1.668' 5.403 3.9823112 0795' 3.743 2.7543116 3.158 2.032 1.901'3117 7.775 10.100 14.5703121 3.612 2.890' 3.7823132 14.375 13.240 10.3633211 1.527* 4.057 3.5823213 4.467 2.833 2.1503220 5.657 2.519 2.6093231 0.484* 1.952* 1.441*3240 2.522 5.469 4.0853311 ó.164 3.747 3.5023320 1.833' 1.167' 0.820'3420 2.981 2.232 1.676'3560 1.118' 2.837 2.20!3693 4.870 3.243 2.4583710 1.958* 1.995' 1.771*3813 4.559 1.175* 1.396*3819 3.769 1.985 1.476'3829 3.818 1.830' 1.572*3843 1.236* 1.987 1.723
which, the simpler one, will he used here. This method used the Kmenta functiotias the null hypothesis and the trans-logarithmic function as the alternativehypothesis. We have:
H0: Knienta function is the adequate modelH1: Trans-logarithmic function is the adequate model
Let Q be the sum of the squared residuals of the regression for the Kmentafunction, Q the sum of the squared residuals for the regression in the trans-logarithmic function. The statistics for examining the null hypothesis is.
(8) F - Q)/2Q1,/IN (K +- I)]
which has an F distribution with 2 and N (K + 1) degrees of freedom. Thedegrees of freedom in the numerator correspond to the difference in the numberof independent variables found between the Kmenta function and the traiis-logarithmic function. The degrees of freedom in the denominator correspond tothe trans-logarithmic function.
An increase in the statistic F in expression (8) indicates a considerable increasein the sum of the squared residuals when the heterothetic function becomes thehomothetic function. In other words, there is a resulting unexplained increasein the variation of the observations when the hoinothetjc function is used. Whenthe statistic F in expression (8) has a significantly small value, it indicates thatthere is no visible change in the explanation given for the behavior ofestablishmeritsin one industry when a heterothetic function replaces a homothetic function.The statistic F in expression (8) has been calculated for each Industry: the valuesappear in Table I I. Negative F values were obtained for five industries; this in-dicates the homothetic function produces better fits than those of the heterotheticfunction, once the respective coefficients R2 have been adjusted by the correspond-ing degrees of freedom. I believe the basic reason for the number of industriesrejecting the homothetic function not being larger is that the trans-logarithmicfunction does not have a good fit for the majority of the industries, as can beseen in Table 11.
Table 11 shows the homothetjc test at the industry level rejected for halfof the industries with a significant level at 5 percent.37The above results (Chow test and trans-logarithmic production function test)suggest that for a majority of Chilean industries, there is no one single productionfunction with constant elasticities which reflects the technological characteristicsof the firms in that industry, in other words, to determine the magnitudes of thetechnological elasticities in industrial establishments each size grouping mustbe studied separately (or, a production function with variable elasticities shouldbe used).
7. SOME REFLECTIONS ON THE METHODOLOGY OF PRODUCTION FUNCTIONSEC0NONIEt-Ric ESTIMATION
This research was started in the belief that the available disaggregated dataon industrial establishments would permit the estimation of production functionsis important to denote that in Griliches' and Ringstads study there is only one industryfor which the honiothetic hypouiesis is rejected In this paper, the number of industries for which thealternative hypothesis is accepted is considerable
612
TABLE 11HoMonunc TFST FOR THE PRouucTloN Fur4(-1 ION Usi i m IRANS-
L(X,AR!i}IMIC lUNCTION
Type of Values for F Degrees of Freedom (Trans-logarithmic
613
* Values of F not significant at 5 percent.Negative values for F which imply that the Kmen;a function has a
better fit than the trans-logarithmic function.
and give "excellent" results. By "excellent" results it was understood that theestimators of different elasticities would have the signs suggested by economictheory and magnitudes corresponding to those obtained in other empiricalstudies: that the values of the estimators would be quite stable for the differentfunctions that were to be estimated; that the statistics t and F would be highenough to reject any null hypothesis which might contradict economic theory;and that the values obtained for R2 would be close to 1.0.
The first econometric estimations contradicted these expectations. When Iexamined the literature on the econometric estimations of production functions,the results obtained generally agreed with the ideal conditions described aboveand drew my attention to the excellent fits shown by the R2 coefficients (betterthan 0.90, generally). The findings of Griliches and Ringstad are one of the fewexceptions to that rule. This study, just like that of Griliches and Ringstad, usesprimary data, i.e., at the establishment level. Why is it then that studies using anideal source of information (from the production function point of vtew) furnishresults worse than those studies which have no possibility of selecting and refiningthe data? The values of R2 obtained in this study are provided lfl the followingtable:
Industry (see expression K) for F function)
3111 7.495 2-169 (13973112 6.492 2- 39 0.6813116 3.693 2-125 0.3523117 12.770 2-623 0.1973121 2.333* 2- 41 0.4 Il3132 15.846 2-548 0.1993211 --0.789 2-220 0.3203213 7.810 2-138 0.4883220 -3.346 2-187 0.1723231 1.046* 2- 55 0.3593240 1.641* 2-136 0.4933311 - I3.358 2-414 0.1583320 - I2.352 2-174 0.1943420 7.914 2 -144 0.5853560 3.424 2-- 45 0.4023693 --0.315 2- 58 0.1883710 0.050* 2-- 42 0.5743813 12.857 2- 80 (1.5643819 -1.446k 2-115 0.4333829 0.528* 2- 91 0.4713843 3.877 2- 76 0.548
LiTABLE 12
NUMBER OF VAlUES FUR R2 FOR EA(iI FORMUlA I ION AM) DyUI ALW-.
Disaggregated data at the establishment level allow the elimination of thoseobservations which provide unreliable information. In this study, where the givendata were supposed to contain a sufficient number of observations at the 4-digitsindustry level, the establishments were subjected to a rigid selection process(see appendix) in which the quality of observations was emphasized over theirquantity.
After this selection process, I proceeded to examine the effects of measuringthe labor and capital variables by different alternatives. The alternative measure-ments for the different variables showed no significant variation among themselves.This had already been predicted by the high correlation coefficients found foralternative forms of measurement for different proxy variables employed. Thisresult indicates that biases introduced by the type of measurement used do notalter significantly the results already obtained. Therefore, I felt that it was irrelevantto calculate the magnitudes of those biases.
Finally, I have estimated and tested several different functions (others arenot discussed explicitly in this study) and all had equally low fits, some with oddsigns for some of the parameters.
In short, the basic information used is taken at the establishment level:only establishments with reliable data are selected; the economic variables aremeasured in several different ways: a variety of functions are estimated: thenumber of observations is sufficiently large: and the number and type of industriesexamined is large and varied. In spite of all these "precautions," the degree ofvariance in the behavior of establishments is considerably greater than that whicheconomic theory can explain.
In light of the results obtained, re-examining the economic assumptionsinvolved in some of the functions (perfect competition and maximization ofprofits) shows that the norm of bad results is also indifferent to the economicassumptions made. It is necessary to take a closer look at the production functionconcept. The series of bad results would indicate that the concept of the productionfunction is not useful for studying the behavior of the firm; but, if the productionfunction concept is irrelevant at the microeconomic level, having already beencriticized at the macroeconomic level, what is its use?
Re-examining the notion of the production function, I see that the essentialcondition of the concept is including only productively efficient techniques. Thus.a process of selection is assumed to have taken place eliminating the inefficienttechniques and leading to a uni-valued function. This basic premise of the produc-
614
Range of Values of R2Functional
models 0.00 0.24 0.25 049 0.50 0.74 0.75 I 00
ACMS 66 35 13 3
Cobb-Douglas 43 44 26 7Kmenta 50 42 21 7
r
tion function concept was not applied in the selection of observations includedto estimate the production functions in each industry.
There are two alternatives to the traditional method of production futjns'econometric estimation for obtaining the production function that would onlyinclude the efficient techniques:38 (1) The engineering approach, where inefficienttechniques are eliminated by monetary cost considerations, and (2) the linear (Orquadratic) programming method where the function is fitted to the data byminimizing the sum of deviations on one side of the curve or on it. A simplervariation of this second approach is Farrell's method, which provides the efficiencyfrontier for all observations.39
8. CONCLUSIONS
Diftèrent production functions exist Jr an industry's establishments
Through the application of the general Chow test I concluded that fordifferent functional specifications it is possible to reject in at least 16 of2l industriesthe hypothesis of no technological differences existing among the different sizegroupings of establishments. There is an increase in the number of industries forwhich it is not possible to employ only one homothetic production function, sincethe fit of the different functional specifications improves.
The use of the trans-logarithmic production function does not give a con-clusive result. The hypothesis that the production function is homothetic can berejected in only 10 of 21 industries.
Finally, the technological parameter elasticities (product-capital, economiesof scale, and factor substitution) obtained for each size grouping are usuallydifferent at each industry level. Therefore, the need to discriminate between thedifferent size groupings of industrial establishments is clearly suggested from apractical point of view.
These results could be used as empirical evidence to support the assumptionof structural heterogeneity within the industrial sector, i.e., technology varies atthe industry level as the size of establishment increases.
The former discussion suggests that caution must be exercised when usingindustry level data. The industry is not a collection of "representative firms," andthe degree of heterogeneity is fairly high. The size of establishment seems to bean important element in determining the economic and technological character-istics of an industrial establishment.
Principal results of production JI4nctions estimation
(a) Value added-cpital elasticity(i) There is no (.imp1e) correlation between the value added-capital elasticity,
fi, and the relative si are of capital in value added, SK. In other words, if S, isused as a value added-capital elasticity estimator, the elasticity would be over-estimated in 75 percent of the cases. The magnitude of such overestimationfluctuates considerably.
3S L. Johansen. Production Functions (North-Holland. 1912). ChapterS.M. J. Farrell, "The Measurement of Productive Efficiency" jfournal of the Roa1 5ta. SOC.,
Series A, Vol. 120, Part 3, 1957) pp. 253-281, This is done in P. MetIer's Ph.D. Thesis, op. Cit.
615
I
/In the majority of industries. 16 of 21, the largest establishments have a
greater value added-capital elasticity than the smallest ones.Size groupings 5-9. 10-19. and 20-49 persons have / elasticities of
approximately 0.40: size grouping 50 99 persons has elasticity higher than 0.50;and establishments of 100 or more persons have fi values close to 0.60.
(b) Economies of scaleEconomies of scale can he observed in 20 of 21 industries, at the industry
level. These economics of scale fluctuate between 10 and 35 percent in 14 industries;26.4 percent (simple arithmetic mean) of economies of scale is observed for the21 industries.
The results obtained for economies of scale are highly questionable,given the odd results obtained at the size of establishment level. In the latter case,there are diseconomies of scale for all size groupings of establishment, even forthe smallest ones employing 5-9 and 10-19 persons. Moreover, when differentsize groupings reveal the variation pattern of scale economies across size classes,some industries present an average cost curve with an inverted U form.
The test for constant returns to scale is accepted for almost 80 percentof the size groupings when estimated separately. But this test is rejected at theindustry level in 14 ot'21 industries.(c) Elasticity of substitution
(1) The ACMS method (Arrow etal.) produces adequate values for the elasticityof substitution a. These values arc consistent with the expectations provided byeconomic theory. Furthermore, these values fluctuate only within small rangesof magnitude.
The elasticity of substitution varies among industries and also among thedifferent size groupings in an industry. At the industry level, a oscillates between0.8 and 1.2 in 14 of 21 industries, and save two exceptions, the industries as awhole have a values above 0.65. No regular pattern of behavior is observed in thevariation pattern of a in the size groupings of an industry. Considering the fivesize groupings, a would show a U shaped pattern as the establishment size increases.The fourth group (50-99 persons) has the lowest value of a, and the largest group(100 and more persons) has the highest a value. However, the largest groupcannot really be compared to the rest because of its much greater size range.As the latter is an open group it is not strange to find there the highest estimatedvalues of a.
The traditional tests of values that a may take are related to values of 0(fixed proportions function) and I (Cobb-Douglas function). The hypothesis a = 0is rejected in almost 70 percent of the cases. The Chilean industrial structure, ingeneral, does not show technological inflexibility. The hypothesis a = I is onlyrejected in about 6 percent of the cases.
3. Firs of different functional specifications, measured by R2 values, are generallypoor for ailfunctional forms
The traditional ACMS specification produces the worst fits, with 87 percentof the cases having an R2 below 0.50. The Cobb-Douglas and Kmcnta functionshave 75 percent of the cases with R2 below 0.50.
616
I
These poor fits, the odd results obtained for the economies of scale, and thegreat instability and fluctuation of different estimated parameters introducedoubts regarding the utility of the production tunction concept at the micro-economic level. However, to defend this concept at micro level and to explainthe problems iiieiitioacd in the previous paragraph, we would have to indicatethat in the econometric estimation there is one requirement which has not beenconsidered, namely the technical efficiency condition that must fulfill the differentproductive techniques.
The fundamental question to be asked is: How reliable are the results andwhat is their use? These results are merely of a descriptive nature and, given theirhigh variability, are not appropriate for suggesting any kind of economic policymeasure. The purpose of the estimators obtained is to show average valuesindicating an existing empirical situation which in some cases is very differentfrom the one predicted by the theoretical model.
APPENDIX
Data used in this study correspond to a four-digit disaggregation industrylevel, ISIC classification. Basic data consists of primary information at industrylevel for the Chilean Industrial Sector Manufacturing Census of 1967.
The 21 industries shown in Table A-i were selected according to a flexibleapplication of the following criteria: (1) Each chosen industry should countwith a "sufficient" number of observations to enable a meaningful econometricestimation in the different size groupings of establishments (2) industries chosen
TABLE A-IISIC Cooa AND NAME OF THE 21 INDUSTRIES SELEC-rED
ISIC Code Name of Industry
3111 Cattle slaughtering preparation and storing of meat3112 Manufacturing of dairy products3116 Mill products3117 Manufacturing of bakery products312! Processing of various food products3132 Wine industries3211 Spinning, weaving, and textile finishing3213 Knitting factories3220 Clothing factories, except shoes3231 Tannery and finishing workshops3240 Shoe factory, except plastic or rubber3311 Sawmills, barracks, and wood workshops3320 Furniture and accessory factories3420 Printing presses and publishing companIes3560 Plastic products factories3693 Cement products factories3710 Iron and steel basic industries3l3 Structural metal products factories3819 Nonspecific metal pioducts factories3829 Machinery and equipment manufacturing3843 Spare parts and accessories for motorized vehicles
617
I
TA
BLE
A-2
NU
MB
ER
OF
ES
TA
BLI
SH
ME
NT
S E
LIM
INA
TE
D A
CC
OR
DIN
G T
O D
IFF
ER
EN
T S
ELE
CT
ION
CR
ITE
RIA
Not
e: T
he fi
gure
s do
not
add
acr
oss
to th
e to
tals
sho
wn
in th
e la
st c
olum
n be
caus
eso
me
esta
blis
hmen
ts a
re in
clud
ed u
nder
mor
e th
an o
ne h
eadi
ng.
---a
--,-
:,
S
Typ
e of
Indu
stry
No.
of P
eopl
eE
mpl
oyed
less
than
five
No.
of D
ays
Wor
ked
less
than
one
No.
of W
orke
rsan
d E
mpl
oyee
seq
ual t
oze
ro
Bui
ldin
gsB
ook
Val
ueeq
ual t
o ze
roV
alue
add
ed le
ssor
equ
al to
zer
o
Pay
men
ts to
Cap
ia1
Fac
tor
less
or
equa
l to
zero
Mac
hine
ry B
ook
Val
ue e
qual
toze
ro
Tot
al N
umbe
r of
Elim
nate
dE
stab
liThm
ents
3111
30
132
30
216
324
3112
I0
141
00
04!
3116
30
127
00
129
3117
100
253
40
23
526
3121
10
024
00
025
3132
430
1334
2()
10
3032
110
00
125
00
112
532
133
00
199
00
020
032
205
00
483
00
04!
532
310
00
310
01
2!32
401
01
200
02
02(
X)
3311
244
022
096
02
195
3320
50
l28
70
I3
2834
202
01
233
01
122
335
601
II
6I
00
6636
932
00
620
00
6337
100
00
120
00
238
13I
00
III0
00
fl238
193
02
117
00
I11
738
290
00
660
00
6638
430
00
630
00
63
TO
TA
L32
824
443
06II
2843
7!
TABLE A-3NUMBER OF OBSERVATIONS tJsFfl FOR PRtMJ(TlON FtN( iiC1 r MAt
Establishment Sue Cla5
should produce more or less homogeneous products and (3) there should beat least one industry for each two-digit ISIC classification.
These 21 industries have 8021 establishments. These establishments weresubmitted to the following selection criteria:
I. Number of persons employed per establishment is less than 5)Number of days worked per establishment is equal to 0.Total number of workers and employees is equal to 0.Book value of machinery is equal toO.Book value of buildings is equal to 0.Added value is less than or equal to 0.Payment to capital factor, obtained as the difference between value addedand total labor cost, is less than or equal to 0.
In most cases 0 does not literally mean iero but reflects the omission ofinformation.
The establishnents that did not meet any one of the previous criteria wereexcluded from the sample, The number of establishments was drastically reducedfrom 8021 to 3650 (see Table A-2).
t should be pointed out that over 80 percent of the eliminated establishmentsbelong to the two smallest size groupings (5-9 and 10-19 people employed).
l In spite of the fact that the Industrial Census should include establishments employing at leastfive persons, there are 328 establishments violating this rule.
619
Type ofIndustry
5 9 10 19 2049 5099persons persons persons persons
1(X) and morepersons
Totallndustr
3W 78 36 39 16 63112 II II 12 8 II 533ll6 29 30 50 20 2 313117 210 249 143 15 12 6293121 9 16 12 4 6 473132 357 131 56 6 4 5543211 22 37 73 36 58 2263213 36 29 44 20 16 1453220 60 4l 43 20 29 1933231 9 13 20 II 8 613240 30 37 31 20 14 4'3311 124 129 97 33 31 4203320 79 47 37 0 7 1803420 48 25 39 4 9 453560 6 ) 19 7 JO SI3693 21 20 17 5 643710 7 6 17 4 12 483813 l9 27 20 8 2 863819 45 30 32 8 6 1213829 22 4 27 21 13 973843 IS 21 19 12 IS 82
TOTAF 1.237 960 847 296 308 3.650
TA
BLE
A-4
CA
PIT
ALS
SH
AR
E A
ND
PR
OD
UC
T-C
AP
ITA
L E
LAS
TIC
ITY
BY
ES
TA
BLI
SH
ME
NT
SIZ
EA
ND
IND
US
TR
Y
Val
ue A
dded
- P
aym
ents
to L
abor
Fac
tor
Cap
ital S
hare
Val
ue A
dded
9: R
egre
ssor
Coe
ffici
ent o
f Equ
ition
: log
1/ L
= lo
g A
+ /3
log
K/L
+ 6
log
L (S
ee T
able
A-7
)
r: S
tatis
tic fo
r nu
ll hy
poth
esis
H0:
/3 =
Typ
e of
Indu
stry
Cilu
Cod
e
Est
ablis
hmen
t Siz
e C
lass
5--9
per
sons
10-1
9 pe
rson
s20
-49
pers
ons
S/9
/9z
S
3111
0.35
!0.
269
-0.6
300.
58!
0.58
50.
020'
0.66
10.
436
- :3
01'
3112
0.48
41.
547
8.24
40,
711
0.11
8l.7
19'
0.65
61.
411
2.18
8'31
160.
685
0.43
7-
1.28
8'0.
716
0.49
7--
0.9
48*
0.71
40.
372
- 3.
455
3117
0.51
20.
227
-6.1
960.
4.68
0.22
6-6
.050
0.45
30.
101
-'.18
431
210.
392
0.34
0-0
.l75'
0.62
8-0
.069
-3.7
070.
698
0.67
90.
089'
3132
-0.6
110.
172
-9.9
770.
741
0.13
5-8
.684
0.79
20.
367
- 3.
455
3211
0.57
50.
915
2.65
10.
689
0.33
5-3
.222
0.59
60.
513
- :1
23'
3213
0.67
50.
559
-0.9
92'
0.68
00.
193
-4.1
490.
684
0.38
8-2
.696
3220
0.38
70.
339
-0.3
74'
0.62
60.
029
-4.7
550.
618
0.16
0-2
.802
3231
0.72
10.
264
-5.1
350.
733
0.20
2'-2
.313
0.62
50.
459
-0.9
27'
3240
0.53
00.
481
-0.8
18'
0.48
80.
236
-2.0
660.
511
0.36
3-
:244
'33
110.
529
0.40
3-
1.82
0'0.
588
0.35
3-3
.35!
0.37
80.
368
- .0
15'
3320
0.38
80.
352
-0.3
58'
0.41
30.
420
0.59
4'0.
545
0.46
8-0
.849
'34
200.
515
0.38
4-
1.31
0'0.
443
0.43
20.
666'
0.48
30.
451
-0.2
17'
3560
0.52
80.
259
-0.8
39'
0.75
60.
755
-0.0
05'
0.62
70.
433
- 1.
576'
3693
0.52
90.
169
-3.7
190.
555
0.40
8-0
.477
'0.
227
0.14
6-0
.500
'37
100.
414
0.51
10.
558'
0.54
00.
333
-0.9
08'
0.60
80.
346
-- 1
.664
'38
130.
361
0.08
3-2
.248
0.50
30.
366
- 1:
458'
0481
0.16
3-
1.60
0'38
190,
531
0.22
3-2
.885
0.57
20.
442
- 1.
284'
0.60
60.
399
-98
0'38
290.
318
0.42
00.
495'
0.35
90.
265
-0.7
39'
0.27
70.
417
1102
'38
430.
603
0.60
0-0
.017
'0.
645
0.39
4-
1.70
2'0.
635
0.12
4-3
498
TA
BLE
A-4
---(
cont
.)
-
': va
lues
non
-sig
nific
ant a
t 5 p
erce
nt.
(a)
Val
ues
not c
ompu
ted
beca
use
num
ber
of o
bser
vatio
ns w
as e
qual
or
less
than
five
.
Typ
e of
Est
ablis
hmen
t Siz
e C
lass
Tot
al In
dust
ryIn
dust
ryC
IJU
50-9
9 pe
rson
s10
0an
dm
ore
pers
ons
Cod
eS
S,
ftt
3111
0.68
00.
690
-0.0
53'
0.66
60.
598
-0.2
89'
0.50
80.
692
2.32
631
120.
774
0.36
9-0
.707
'0.
670
0.54
8-0
.667
'0.
652
1.01
925
6631
160.
704
-0.3
02-5
.427
0.35
4(a
)(a
)0.
701
0.46
5-
2.98
731
170.
461
0.61
60.
861*
0.59
50.
853
2.66
00.
482
0.29
9--
7.03
831
210.
783
(a)
(a)
0,70
81.
903
1.30
6'0.
624
0.28
0--
2.89
031
320.
817
0.58
0-
1.65
8*0.
675
(a)
(a)
0.66
303
63.
8.33
332
110.
563
0.33
71.
694'
0.54
50.
351
-3.9
950.
591
0.49
5--
2.06
532
130.
577
0.66
40.
475*
0.63
00.
274
-2.7
010.
660
0.50
9-2
.437
3220
0.61
50.
172
-2.1
500.
582
1321
1.97
7'0.
362
0.40
60.
593'
3231
0.53
10.
507
-0.0
47'
0.49
4(a
)(a
)0.
628
0.44
1--
2.3
4032
400.
481
0.35
0-0
.830
'0.
480
0.47
8-0
.174
'0.
500
0.52
10.
356*
3311
0.46
30.
018
-3.6
560.
350
0.38
60.
274'
0.49
10.
44)9
--2.
222
3320
0.36
00.
939
3.21
80.
482
0.57
20.
236'
0.42
90.
445
0.27
8'34
200.
400
0.64
91.
618'
0.43
80.
601
1.24
2'0.
473
0.62
53.
118
3560
0.63
40.
994
1,23
9*0.
634
0.08
4-2
.170
'0.
651
0.48
7--
1.68
9'36
930.
553
(a)
(a)
0.46
9(a
)(a
)0.
458
0.31
3_.
l.918
*37
100.
641
(a)
(a)
0.51
30.
366
-0.7
93'
3.31
70.
547
0.00
0'38
130.
529
0.56
60.
145'
0.34
00.
253
-0.4
09'
0.44
70.
338
-. 1
.697
'38
190,
559
0.24
8-
1.27
5'0.
509
0.50
4-0
.042
'0.
562
0.41
1-
2.64
238
290.
496
0.39
4-0
.390
'0.
419
0.55
31.
139*
0.36
50.
559
2.70
238
430.
556
0.56
40.
053'
0711
1.01
92.
217
0.63
40.
506
- 1.
803'
TA
BLE
A-5
EC
ON
OM
IES
OF
SC
ALE
AC
CO
RD
ING
TO
CO
BB
-DO
UG
LAS
AN
D K
ME
NT
A P
RO
DU
CT
ION
FU
NC
TIO
NS
BY
ES
TA
BLI
SH
ME
NT
SIz
E A
ND
IND
US
TE
Y
Hc_
0: C
obb-
Dou
glas
eco
nom
ics
of s
cale
. (T
able
A-7
)h:
Km
enta
eco
nom
ies o
f sca
le. (
Tabl
eA
-8)
Est
ablis
hmen
t Siz
e C
lass
'h v
alue
s w
hose
t st
atis
tics
in n
on-s
igni
fican
t at 5
per
cen
t. S
ee T
able
s A
-7 a
nd A
-8.
(a)
Val
ues
not c
ompu
ted
beca
use
num
ber
of o
bser
vatio
ns w
as e
qual
or
less
than
five
.
Typ
e of
lndu
stty
CII
UC
ode
Tot
al In
dust
ry5-
9 pe
rson
slO
--19
per
sons
20-4
9 pe
rson
s50
-99
pers
ons
100
and
mor
e pe
rson
s
h.0
h,,
h.0
/Ic.4
)h1
c_D
3111
-0.6
56-0
.660
-1.3
50-1
.354
-0.1
74'
O.l8
l'0l
53'
-0.2
47'
0.74
8*0.
363*
-0.1
19-0
146
3112
0.14
8*0.
040*
0.39
0'0.
469*
0.48
8*0.
194*
-0.9
92'
0.69
9'-0
.471
'00
49'
0.29
40.
308
3116
-0.2
93'
-0.3
450.
125
0.17
7*04
95'
0.53
01.
062
1.05
8(a
)(a
)0.
263
0.26
031
17-0
.707
-0.7
13-0
.857
-0.8
21-0
,140
'-0
.380
'0.
236'
0.32
40.
243'
0.24
1'0.
080
0.80
031
210.
787'
1.00
3'-0
.191
'0.
190w
0.21
80.
473'
(a)
(a)
1.22
8'1.
141'
0.45
40.
461
3132
-0,4
00-0
.396
-0.3
74-0
.374
0.17
7*0.
294*
0.15
2'0.
139*
(a)
(a)
0.06
200
65j
3211
-0.6
13'
-0.5
97'
-0.2
67'
-0.3
95'
--0.
575
-0.5
66-0
.142
'-0
.145
'--
0.02
3'-0
.022
'0.
070'
0007
'1'
.)32
13-0
.249
'-0
.104
'-0
.747
'-0
.147
'-0
.353
'--
0.35
0'-0
.668
'-0
.636
'-0
.007
'-0
.006
'0.
216
017
3220
-0.2
77'
-0.3
31'
-0.7
38-0
.779
-114
04'
-0.2
86'
-0.3
54'
0235
'0.
061*
-0.1
25'
0.25
502
3832
310.
233
0.28
9'1.
087
- 1.
044
-0.1
20-0
.155
'-0
.404
'-0
.499
'(a
)(a
)0.
128
0.13
832
40-
1.40
3-
1.36
5--
1.18
0'-0
.184
'-0
.513
-0.5
50-0
.289
-0.4
510.
265
0.25
70.
291
0291
3311
-0.0
15'
-0.0
23'
-0.1
58'
-0.0
17-0
.124
'-0
.119
'0.
082'
0.01
7*0.
184'
0.18
1'0.
060
0550
3320
-0.5
2'-0
.084
'-0
.279
'-0
.303
'-0
.363
'-0
.298
'0.
000
-0.0
16'
0.58
5'0.
261
0.29
903
1034
20-0
.791
-0.7
93-0
.680
'-0
.662
'-0
.311
'-0
.318
'-0
.967
0.02
7'0.
264'
-0.2
66'
0.02
1'0.
021'
3560
-0.9
72'
-0.9
80'
0.32
2'-0
.281
'-0
.434
'-0
.440
'-0
.436
'-0
.866
'0.
453'
0.45
7'0.
088'
0.09
536
93-0
.532
'-0
.551
'0,
864'
-0.8
55'
-0.8
71-
1.21
7(a
)(a
)(a
)(a
)0.
157'
0.16
2'37
10-0
.103
'-3
.621
'-0
,325
'-0
.880
'-0
.455
'-0
.533
'(a
)(a
)-.
1.05
0'-0
.185
'0.
118'
0115
'38
13-0
.421
'-0
.199
'--
0.24
9'-0
.425
'-0
.261
'-0
.257
0.22
4'0.
183'
0.20
4'-0
.009
'0.
343
0334
3819
-0.5
56'
-0.5
80'
-0.3
01'
-0.1
93'
-0.1
26'
-0.1
96'
-0.1
29'
-0.2
93'
0.16
1'-0
.047
'0.
318
0326
3829
-0.4
95'
-0.4
70'
-0,8
68'
- 1.
383
0.31
1'0.
3 12
'-0
.183
'0.
076'
0.41
70.
457
0.23
002
2438
43-0
.152
'-0
.281
'0.
623
-0.4
61'
0.52
2'-0
.549
'-0
.493
'-0
.580
'0.
035'
0.07
1'0
353
0353
Typ
e of
Indu
stry
CIIU
(T
odc
TA
BLE
A-6
AC
MS
PR
OD
UC
TIO
N F
UN
CT
1ON
EC
ON
OM
ET
RIC
RE
SU
LTS
BY
ES
TA
BLI
SH
ME
NT
SIz
E A
ND
IND
US
1RY
I'rod
uctio
n F
unct
,on
1' -
y)4
K+
II -
Reg
ress
ion
Eqt
rnt,o
n: lo
g Y
,'L -
a 'I
- cl
og
Esi
abI,s
hrnc
nt S
ize
Cla
ss
5-9
pers
ons
10-1
9 pe
rson
s
R2
0
70-4
9 pe
rson
s
NJ
3111
-5.7
9194
50.
250
-5.7
401.
598
0.33
4-3
7)0
0.90
101
:
-.2
59)
(516
2)(-
9.45
4)(4
308)
(-8
165)
(7.4
34)
3112
-5.8
5515
990.
338
- 5.
178
.205
0336
-6,5
771.
943
0 50
9
- 9.
169)
(7.4
7))
(-6
740)
(2.5
08)
1-7
086)
(4(7
6)
016
-6.0
44-1
0.16
6)72
))
(424
7)03
79-.
3795
I-.
7.4
93)
0.59
0(2
001)
0094
- 3.
439
(-77
77)
0467
2.20
1)
0,07
3
3117
-5.4
2407
3606
4-5
327
0779
0203
-489
606
080
(75
( -
35.7
40)
(694
0)I
. 37.
5871
(802
0):7
.s:)
7)(4
.680
1
-S
I')-9
72)
08)9
0335
-3.9
0604
880(
2))
-5(9
91(
1,8
0454
(-63
39)
(224
21..(
( 64
0)I)
747
)-6
661)
(364
)
-3(
5'.3
529
--02
8401
18-3
034
-027
00)
02-2
.893
0.16
10.
0)3
(-65
74')
-696
2)(
-44.
366)
I-.
1963
)-(
7547
)((
35(1
)
-3'))
-S 3
25(.
182
(124
)-4
.263
0.76
6(3
247
-444
307
7707
(4
-. 7
.854
)2.
7(37
)I -
II 7
58)
(357
7)(
- (3
007'
(4.3
97)
92)3
-5(3
40.
89)
(((5
))-
45)1
507
(10
55-
3.77
604
33((
051
I- 3
460
775)
)(-
(2.7
70)
(247
6)10
501)
((8(
6)
-3'
70-4
854
0.59
100
72-4
.24)
043)
0025
-477
809
:)1(
3:0
723)
( -
15.0
80)
-432
2
2.35
7)
0634
05(2
(10
.484
)
-4.3
22
((42
(3)
0)15
90)
160
- (5
.756
)
-434
6
1454
1,
(('4
60
17'
(-13
,070
)30
63)
(-5,
543)
(I 3
27)
-599
3)(1
9)5
)
3740
-495
80.
527
000)
1-5
066
0907
(373
4-5
365
(20,
903
)16
(-8.
753)
(1(0
8)(-
-(4,
458)
(320
01I.-
((0
37)
(376
9)
.311
-4(1
0-4
2,52
31-0
,034
(-0
382)
--0.
007
-3 5
04I -
3902
8)-0
.153
- 2.
1)90
)00
26-4
,071
(-73
299)
021(
1,
(225
0100
44)
3370
- 3.
304
I 0)9
0.25
)--
5231
0,87
30
153
--42
9904
520
(0)
1-22
.500
1(5
212)
I -(2
,713
)(3
032)
I -14
.308
)i2
746
)
r
TA
BLE
A-6
---(
coni
.)34
20-5
704
- 16
.614
)03
6(5
,646
)03
97-6
.452
( -9
.194
)1.
52)
(449
1)0.
444
4.49
3-9
411
0699
3267
0203
3560
-4.3
(2(-
4.59
4)0,
404
-0.4
6-4
.46
0765
0.20
6-4
.059
0 50
400
71)
3693
--4.
641
(-13
506)
(0.5
03)
0.30
800
28
(-6)
06,
--4.
973
(I 7
60)
0654
0.50
6
(-60
50
-343
6
I93
3)
-0 3
0-3
037
((.2
54)
(-(7
.667
)(4
526)
(-49
)5)
-077
5)37
)0-5
.101
- 3.
938)
0.60
8-0
.094
--5.
037
01)8
305
15-4
60)
0956
3320
0.69
7)(
- I
400)
(2.9
06)
-743
8)2
967)
3813
-589
15.4
69)
0462
0 39
-504
90.
830
0204
--3
06I 7
34(
-(1
774)
-- 0
.995
)2.
767)
-13
IOn
5 2S
5
38(9
--5.
199
18,3
74)
0.1)
320.
243
-4.5
5)05
470)
2)-4
055
0485
'I-
3.44
6)(-
3046
)(2
234)
(-09
3')
(274
6)38
29-5
.933
10.4
701
0885
0223
-3.3
3-0
.097
-0,0
77-5
207
0946
3 21
1'(2
.65)
)-
4.57
9)(-
0.26
0)-8
921)
(3)8
8)39
43-4
866
(-10
.329
)0.
855
0 27
8-
4.27
20
696
0.4)
- 7
793
0 50
9.1
082
(252
5)(-
7.60
4)2.
070)
(-65
3))
(6(4
)
EsL
abI,s
hn,e
n, S
izr
C)a
ss
Typ
e of
lndu
5try
Cli)
) C
ode
50-9
9 pe
rson
s00
and
mor
e pe
rons
lola
) )n
4us'
r,a
RI
aa
ap
0.12
1)67
5(0
3)7)
- 02
50(-
1846
9)(7
690)
25)
3)12
(-20
23)
-006
4(-
i.H7)
-0.1
(11)
1)8(
9)12
80C
604
31)6
- 3.
640
0.59
90.
259
(a)
((-
7.12
))(2
762)
45)
0935
0305
3)17
'121
:, (a)
2.04
2)
(a)
0)64
(a)
-9.3
24
1818
8)
3)85
0762
190_
9,
-593
92)
-508
0
(762
4)
1, I0
021
(I 4
))
3132
- 2.
554
0.06
0-0
.204
(-4.
796,
(a)
(3,7
77)
I
1-14
124
(3
99)
(-9.
269)
( --
0.37
4)-3
319
-0)0
711
913
-90)
07)
-301
)3
p
L
Val
ues
in p
aren
thes
es 0
rest
atis
tics
br n
ull h
ypot
hesi
s th
ai th
ose
para
met
ers
have
a z
ero
valu
e.
(a)
Val
ues
not c
ompu
ted
beca
use
num
ber
of o
bser
vatio
ns w
as fi
ve o
rle
ss.
3211
-3.3
0)-
S 6
60)
0257
(0.8
68)
-000
7-4
,677
(- 1
4.58
4)0.
8)10
(5 5
74)
0.34
1,-4
.476
- 25
.297
)0.
820
(860
7)0.
241,
3213
-318
2(-
4776
)-0
.003
(-0.
005)
-0.0
56-2
.464
) -
3.48
81-0
,11)
11-
0.37
.9)
-0.0
63-4
796
(-27
126
)0.
936
(6.0
7))
0 30
0
3220
-028
81-
5959
)0)
240.
35))
-0.0
47-7
.587
(-17
.330
12.
403
(10.
733)
0801
-5(3
7I -
28.
554)
(070
(II 9
72)
(17.
93
323)
- 3.
452
0.39
2-0
.042
(a)
(a)
a)-4
.356
(16.
1,14
)
0743
(5.5
55)
0.73
2
- 3.
1 36
)(0
.776
)
1240
-4)2
8-9
260)
0595
1251
2)0.
2(8
- 4.
589
(-9.
41(0
)0.
679
(364
7)03
48-
5.42
2I -
30.
045)
I 206
(103
7.8)
((.4
28
3311
-4,1
56(-
16,8
66!
0 36
0(2
.393
)
0 29
- 5.
324
- 9.
232)
040
(6.3
13)
0.5)
9-
3.59
4(-
6444
4))
0078
(169
6)00
03
3320
- 3.
580
I--
L690
(00
60(0
052)
-0)2
5-
7.18
51-
5.5)
3)2.
203
(109
))0.
588
- 5.
353
(-3!
471
).0
0>
>8.
330)
0276
3420
_805
!-
8.04
2)
2.01
9(5
227)
0.66
9-
3.20
01-
3439
)0
690.
466)
-O 0
46-5
61!
I -24
1,7.
!)I
iOO
10.5
121
0432
-0.
391
-0.1
45-5
056
I01
,10
749
-436
40,
75(1
(308
3)(-
2275
)(0
492)
(-II
79))
(528
0)1-
12.6
14)
3693
(a)
(a)
(a)
(a)
(a)
(a)
-467
.5(-
2271
)1)
06)6
481(
71
0 21
41
37)0
Ia)
a)(a
)-4
.9(5
..785
'l09
00(3
.535
)(>
511
- 5.
4(7
I - 1
8.12
2)I >
9680
th))
3813
-5)4
5(-
5480
).0
632.
168)
0.34
6-4
.030
>64
94)
0.50
6(>
82),
0)74
-5.5
(1(-
24.8
(4))
>02
8473
)((
455
0619
-490
>09
290.
106
- 5.
875
1 35
50.
604
- 5.
145
((.9
5383
20)
1(36
7
(-36
34)
(135
3)(-
-6.5
42)
3,43
9)(-
786(
111
3829
-548
8.2
300.
159
-4.9
950.
856
0230
--38
061(
30 8")
(7(-
4278
1(2
.183
)(-
5.37
3)(2
.) 3
9)-
17.1
90)
3843
-449
00.
6>8
0246
-5.7
5315
4404
3>1
-4.6
17(.
042
>84
)
(744
9
(-54
69)
(2.1
44)
(-5.
161)
345!
)-1
7789
)(
TA
BLE
A-7
CO
BB
-DO
UG
LAS
PR
OD
UC
TIO
N F
UN
CT
ION
EC
ON
OM
ET
RIC
RE
SU
LTS
BY
ES
TA
BLI
SH
ME
NT
SIZ
E A
ND
IND
US
TR
YP
rodu
ci,o
n F
unct
ion
Reg
ress
ion
Equ
atio
n5
ILK
'In
V/L
In .4
+ l
+ /3
-I (
is L
-./3
InK
IL
Est
ab(is
hmcn
s S
ue C
lass
Typ
e ol
I nuj
usu
r
CIII
.' C
ode
59 p
erso
ns0-
19 p
erso
ns30
40 p
erso
ns6
/31,
/1R
'p
p1
(a)
(a)
al31
11-0
.656
- 5.
374)
0.26
92.
087)
0.52
5-1
350
1-4
9â0)
0.5i
4(2
.944
)05
06-O
l'4-0
739
)'04
3625
24)
02(1
3112
0.14
8)°
.298
r(5
470.
936
0.39
00
(18
-062
0488
1441
06(1
1;(1
1.46
1)10
77!
)'(0
347
)'0
766)
'49
021
3116
-0.2
93I -
0.42
3)'
0.43
7)0(
310.
125
0.49
71(
550
-049
507
7202
7:(2
62)
2.19
0)(2
.151
)I 8
87)'
(3 '5
9131
17-0
70)7
-5 0
72)
0.22
70.
254
-0.1
)57
0.22
602
00-0
140
0(0)
0041
(4.9
33)
- 6.
707)
(559
6)-
014)
'(2
054)
3(3)
0.78
7(I
3)3
)'05
403
-003
301
91-0
.069
-01(
40.
218
0679
(06:
(.16
4)'
(055
7)'
(-03
69)'
(046
2)'
(286
813j
32-0
.4(
- 7.
5%)
(1.1
72
(3.9
22)
0284
-0.3
74(-
4.10
4)01
351(
935)
'02
4901
770.
654)
'03
6729
77)
0(11
'
3211
-0,6
13-I
575
)'-0
915
0774
-076
70.
335
0271
' 551
505
)304
6017
136)
1-0,
768)
3 04
61I-
0078
)Ii,
95)
132
(3-0
.249
(-05
09)'
0.55
933
60)
0291
-074
7(-
0.23
9)'
0193
(I 6
52)'
1)03
2
--I 9
12)'
0 36
810
5501
0 M
3231
)-0
.277
(-O
607
0339
(264
')0.
177
-073
91-
2546
)0.
029
(022
9)'
0(20
-040
4I
I(3
5)'
OIo
O10
9761
'05
82
03(1
0233
10 6
07)'
0264
(2,9
56)
0486
1.0)
87
(-'3
455)
0 20
2(0
.880
)0
1,2)
-0 2
0(-
0292
)'0
459
1:56
11C
I
3241
1--
I 406
- .1
5(1
9)04
31(3
.562
)0.
537
--0
180
I-0
481u
'02
3611
934)
'51
)75
.-05
13(-
-237
))03
6.1
,S 0
34)
749)
3311
00(5
I -0
75)'
0403
5797
)02
67-0
.158
(-2.
222)
0353
(507
610(
30-0
524
I .-0
405'
0368
(372
5)(1
(15
P
00
TAB
LE
Typ
e of
Indu
stry
clut
'C
ode
A-7
----
(cgn
i.
loca
l Ind
oory
Fst
abl,s
hrne
n, S
ire C
lass
50-9
9 pe
rson
s10
0 an
d m
ore
pers
ons
itH
/1R
'ii
18'
3)))
31)2
7116
3117
3)32
12)1
3213
722)
)
323)
1240
33)!
3320
-0.1
531-
0.42
4)'
-0.9
96-0
892)
'
062
7.18
01
0.23
8(0
.38)
)'
13)
052
(1.1
)68)
'
-0.1
42-0
705
)'
-068
8(-
1.75
0)'
-035
4-1
000)
'
--04
04-0
84)
)
--02
69-
1284
)'
0082
(070
))'
0002
0008
1'
069!
)(3
69(l)
0.36
9(0
.645
1'
-030
2-1
.632
1'
06)6
(342
0)
14)
0,56
036
04)
-033
7(2
509)
0.66
4(2
783!
0.17
2(0
815)
'
0,50
70.
986)
'
0350
(722
21
0.0)
8l0
)45r
0939
S. 2
09)
0469
-004
3
0)40
0.36
)
1)
0685
0)26
045?
0024
0057
0374
- 0)
198
0759
0749
(262
3)'
-047
!-
-
- 70
)'(.
1)
0243
(112
4)'
.228
(2)9
!))'
IS!
-0.0
27(-
0.26
6)'
-0,0
07(-
0063
)'
0.06
1
(ii IS
)'
)
0265
(2.2
98)
0.18
4II
023)
'
0.58
514
76!'
0.59
%(2
539)
'
(1$4
0-.,
2998
)
(a,
0,82
38.
453)
1903
(200
11'
(15)
1)S
I(3
149
)
0.27
4(2
.076
)'
32!
7573
!
a)
0478
4,60
0)
4)18
6(2
,928
)
0.57
2,)
0 66
0
,s..s
,
() 0885
0783
022
0.17
9
0274
(a)
0528
0.16
0
0378
I 4$!
7 89
7)
I --0
975
)'
-- 2
995
- 29
)!)
-370
4--
1' .8
2,
--59
4)(-
-9 5
64)
'99
-- 1
0.31
31
--06
72-
I55
3)
- 28
29I
- 47
66,
--36
44
-197
7-2
(27)
5 22
!
-52
83!
- 21
3!(-
8 32
41
-787
9--
6003
)
-(11
19(-
2.11
9)
094
Il 03
3j
0.26
312
8331
0090
(2 1
9$)
(1.4
54
(4.8
191
0)16
2I 9
67
0070
0207
1'
02)6
14.6
74)
0.25
54
('6)
0)28
(155
2,'
0.29
)(6
029)
0060
(225
6)
0 29
914
774)
('092
($74
71
10)9
7 (2
0)
0.46
557
9!,
0299
fl 5)
71
0290
(250
7)
0 63
1,
(9,9
32 I
0.49
5((
0637
)
0709
8)96
!
0406
1541
2)
044)
(551
81
(152
1
1) 9
)8)
0409
II) 0475
1771
5)
a)
0343
08)4
03(4
0 79
0 39
0137
((33
5
04%
)
02).
)
037
0491
1
Ii 05o
a
valu
es n
onsg
nitic
aflt
at 5
per
cen
t.ti)
Val
ues
not c
ompu
ted
beca
use
num
ber
of o
bser
vatio
ns w
as(I
ce o
r le
ss
3420
0967
71,4
90.
744
0 26
4l.1
1117
)
0.60
1(4
,570
)
0.53
50(
4511
0,02
1(0
,467
r0.
625
(127
611
054%
(-2.
270)
(421
))--
3560
3693
0436
- 0.
1153
)
(a)
0119
413
4111
0696 (4)
0.45
3.1
51
(a)
0.06
403
2$)
(a)
-.00
70 al
- (.
417
- .5
17r
1440
2%3j
(
0.01
,11
II (5(5
1
0.41
17
(5.0
121
1(31
3(4
.1 2
(11
1334
3
(172
8
37(0
(4)
(a)
1.05
0(-
0 5l
0(0.
366
(I 9
65)
0(74
(563
- I 5
531
0118
0547
1159
3
38(3
0,56
60.
291,
0 20
40.
855(
0.25
3(I
184
)00
42--
4955
(-59
521
0 34
3(4
741
1)
1)33
815
.235
1
0436
1031
012.
2l9)
03(8
0411
0 44
111(
9
3629
3843
0 (3
90.
178)
-0(1
13o.
28)r
0.49
30.
981,
)
0.24
8(1
.01
0.58
4i.s
osr
0,51
,4(4
.082
)
--0
49
0.05
8
0,57
1
(1.1
61
0 51
(2)
0417
(3 (
(1,1
0,03
50.
1 60
1
0.50
4(0
.532
(
0553
(466
2)
019
(7,3
'19(
071,
9
01,7
4
((78
3
251,
7-
3356
)
4,11
731
0.23
0((
1,90
1
(135
3(5
,3 7
8)
I'64
1
0.55
976
09)
0511
67
1161
0475
(151
'
I
TA
BLE
A-8
KM
NT
A P
RO
DU
CT
ION
FU
NC
TIO
N E
CO
NO
ME
TR
IC R
ES
ULT
SB
Y E
ST
AB
LIS
HM
EN
T S
IZE
AN
D IN
DU
ST
RY
Pro
duct
ion
Fun
d on
. YIll
I-
S)L
-R
egrc
saio
n E
quat
ion
log
YL
log
B i
1-
I) lo
g L
+ p
O lo
g K
'L *
alo
g K
. L1
Val
uen
in p
aren
thes
es a
rcS
lalis
lIcs
for
null
hypo
thes
is th
at th
osc
para
met
ers
have
a z
ero
valu
e.
1;
T,C
of
Indu
stry
CIII
.! C
ode
stab
lishm
rnt S
ize
Cla
os
5-9
pers
ons
10-1
9 pe
rson
s20
-49
pers
on..
600
R0
6
311
-0.6
60-
3591
0025
0,04
1-
0.02
4I
- 0.
412)
'05
19-
1,35
4(-
49)4
)-0
760
1-03
64)'
-0(7
5(-
0.14
7)'
0500
-0)8
)(-
0.79
))'
-(02
4(-
1165
)'-0
.162
1-)
695)
'0
253
3(12
0.04
0(0
093r
3.03
802
120.
942
0.46
92.
270
0.20
9-0
.242
0.19
4-(
730
-031
2(7
722)
(157
2)'
(0.5
79)'
(0 '5
7)'
(072
3)'
(024
K)'
) -
0319
)'(-
0 v9
41'
316
-0.7
45(-
4,05
0)0.
337
8.16
71'
-0.0
09-0
.050
)'00
950.
177
(030
5)'
-1.0
67(-
0.58
5)'
-0)6
4)
--0.
719j
'00
70--
0530
(-20
70)
--09
78(-
1019
)'-0
149
(.
I
075
.117
-.07
(3-5
3511
(111
2
(0.2
99)'
-0.0
09I -
0 30
6)'
0,53
4-0
521
- 6.
423)
0 90
6(2
.764
)00
52(2
.086
)(1
.394
-0 3
90(
--I 0
29)
0 11
77
(2.0
42)
0047
I
-0 (
83.5
3121
(.00
3((
.269
)'1.
895
0)53
-0)9
40.
190
0.56
30.
062
02)6
0471
4327
0.15
7((
3(0
.573
)'(0
472)
'0.
530)
'(0
.262
)'(0
.296
)'(0
948)
'(1
500)
'II
268(
3(32
-039
6(-
7587
)-0
1)7
(-05
(0)'
-003
5(-
1315
)'02
96-0
374
(-4)
74)
-006
8(-
0.18
7)'
-002
3(-
0.55
6)'
0245
((.2
94
(3)5
)'-0
)6)
-039
))'
-009
0(-
1342
)'0(
10
321)
-0.5
97-
(.66
9)'
2.11
790)
9608
(54
-0.3
9506
750.
046
0279
-056
616
750(
1400
16(3
157
)(2
)71)
(-10
09)'
(.3l
r11
024)
'(
- 3.
043)
1240
1)II
675'
3213
--0.
104
1-1)
19)1
)'(1
130
(I0.
089
0.27
8-0
147
-1.5
57-0
.152
0092
-035
008
0900
4502
1689
)'(0
.787
)'(-
04)1
1)'
(- (
.487
)'(-
.650
1'(-
(87
5)'
(081
))'
(043
6)'
3220
-0.3
21(-
0.75
3)'
--0.
035
(-0.
032)
'-0
030
(1)0
0-0
.7'9
-0.4
56-0
.045
009)
-02n
632
0502
7200
42(-
0 34
9)'
(-2.
567)
1-0.
402)
'(0
597)
'(-
0762
)'1(
679)
'((
607)
'32
3102
8Q((
1633
)'04
41(0
755)
'00
16(0
707)
'03
09-1
044
I -30
52)
I 901
1
(II 1
125)
'0(
49(0
.577
)'05
77--
0155
(-0.
36!)
'0.
577
I - 0
.265
)'-0
)00
-0.4
77)'
0s
3240
- I
3115
- 3
3'2
(343 .0)9
)'00
68(0
.695
r05
)9-0
94
(-0.
4113
)'-0
.769
(-0.
297)
'-0
.053
) -0
.406
)'00
46'
.Ucc
1)
-234
6)(0
914)
'(0
S77
I04
53
.o___J. S.cY
t
.sIIS ,xaw
iso poniiiluo,)
.1)900).16660)
6)00£Q
.198') 0)I -
6990-0Z
i)-
#M00
)LI'90 -)co-
(L):60)890
lcs9O1.600
001.)190:0-
6986
Z 0
0p 0
.iOp0)
£00
.0)) I)
.(p6r01
91.)
6)
.(co8 0)tI). 0
.19890 - )9610
L):LO
£P0
1.9) 1)
)):6c1)6910
100)):)I1S
IZ
)pZ6l)
lI'L):
i6S90
£610-
c0
06)0
.18') I0910:0
(zc0-)
1900-
.101.10)Lcr 0
.16910-99)0-
0LP'0
.)):OL1 -)
0890-
61:8):
61))):
0).(coi)
(c91c)
11'L0 I)l1.S
I)II--)
0900-.1090) -1
8I10.)69&
0I
LIt'I.(LI9 0
66)0I8
co
6090
189-0
),I)):Cj
icvo
c9 10
6O
I .. 0-:; 0-
9)):06 0
.)LpcO)
.19890-->
.R 1.91)
.0)8 0)9)0:0
01L-0)
ito)L0)) Z
I
.)6Z0)
I)) 0
.19600-)O
ZL0
)0tZJ
999 1
.1961.0)
.)ocZ--
(IL): £ -L1
1-
.(9L): - I0990
(0)z I-)
.1):):):I -
6Z0
.191.80-)6110-
90c900
0t'0
6Z9 0
LZ90
I00
)9EO
Z>
l00
.166)0-)oo-
()'I0)9100
.)OZ
i I01.60
.(Z6' 0-)
9901.-
.(99c 0-)I600
.lc#60-)L(00
)IZ 10
1.690-
.1)99 0)6990
.(pcw1)
89):0I
0c0:0)£900
I
L600
6600-)Z
PO
0
.IOt'8'( -
0820
.ZZ
Il>9980
.)ct0-)
.1)-
Z990
.1891.0-)9090--IO
LIG-
1)990
Z0I 0
)800
9990
9t'I 0
1.860
.16101)iI0
.l:c60)t'IO
O
.1910:0-)9100
.(pOrI)
800
.160:60-)0I00
.)ZI III
691.9
.19990)0610
.111.60-)-0)
910
.19 9 Z)
Pc):)
.)980)Z
6Z0
19660-)1Z
99
.1669 I)Içc0-I --£61. 0
.)8t'zO-)
98000-)61:00-
0)1.9
9690
0986
01:9):
II):):
TA
BLE
A"8
--{c
ont.)
05
Tvp
c of
Indu
slry
C!IL
J C
ode
slab
!ishn
ier.
i Siz
e C
lass
Tot
al In
dust
ry50
-99
pers
ons
00 a
nd m
ore
pers
ons
It8
ItP
S31
11
3112
3116
3117
3(21
31.2
3211 2I3
3220
323)
1241
)
3311
3120
3420
-0.2
47
-0.6
99(-
0.49
1)'
(.05
8(2
.106
)
0.32
4(0
.789
)'
(a)
0 39
(1.4
01)'
-0.1
45(-
05(8
)'
-063
6-
I 359
'
0.23
5(0
.461
)'
-049
9-
I 093
)'
-0.4
5)--
2.09
3)
0.11
7(0
4) 1
)'
-00(
9(0
046)
'
0.02
7(0
.049
)'
- 3.
523
- I 4
)4)'
- (0
.605
1-0
486)
'
0.00
60.
003)
'
3.88
9(2
9)1)
(a)
1064
(0.6
84)'
.440
(1.7
09)'
-53
01-
0 26
4)'
- 6.
981
) -
I 635
)'
-114
39)
I67
)'
4464
(234
2)
-0.4
05-0
4(6)
'
270
(0 4
97)'
06))
)(0
38,'
--04
43)
- l.6
45)
--1.
304
I - 0
.504
)'
0.03
7(0
(69)
'
0.35
7(2
,456
)
(a)
0.06
0(0
.327
)'
0158
((07
3)'
0.23
0-0
378
)'
-0.8
65--
(.6
80)'
- 1.
369
)-
7)5)
'
1136
9
2)60
)
-0.0
41(0
.44)
)'
206
(028
3)
(1.0
03
(001
3)'
0331
-0.4
)9
0.09
9
9)72
2
(a)
0.88
4
0,14
0
0.21
1
-0.4
90
0(73
0462
--00
85
0907
0 52
6
0.36
3(1
.430
)'
0.04
9(0
.136
)'
(a)
024!
(1.2
00)'
I14
1
(3.1
96)'
(.t)
-0.0
221-
0.35
5)'
-000
6-0
063)
'
- 0.
125
--0.
340)
'
(a)
0.25
72.
207)
0.18
1(0
.999
)'
0261
(0.5
04)'
-- 0
,366
10.1
47)'
-9.3
26(-
2.07
3)'
5.43
02.
656)
(a)
1393
1300
1'
4905
0(1
639)
(a)
0.15
1(0
.242
)'
--01
(1-
0.05
7)'
-- 1
0.13
8-2
,469
)
Ii)
-045
6I -
-039
))'
I ((9
)((
.0(1
1)'
0060
(101
)4)'
0 19
0II
770)
'
-0)0
41-
2 20
8)'
057)
(2.3
95)
(a)
0.05
30.
528)
'
5 (0
5(1
574)
'
a)
-002
2(-
0.16
9)'
--00
40(-
-0.1
83)
--0.
084
--28
00)
Ia)
-006
6(-
0.90
4)'
0.06
410
.648
)
0794
((.0
23)'
0044
(0i8
9)'
0.10
9
06 9
0 0.90
)
0906
(a)
0.11
)9
0430
043.
1
(( 0526
0.14
1
0366
9) 4
9)
--04
51-0
390)
'
0.43
80(
58)'
-- 5
.449
-323
0)
-109
5--
.550
)'
- 4
007
--I 8
48)'
-3.0
9)(-
7068
1
-0.6
13I
- 0.
949)
'
_355
- I
--52
37I
-312
1,
-.I 4
46-0
.905
)'
-461
17--
2799
)'
--39
70(-
-4 I5
(
-216
2--
I ((05
011
0971
'
-0(4
6)
-253
3
(130
6(7
.097
)
(028
0(9
3833
)
0081
)2)
98)'
0461
4751
)
0085
(2 0
47)
0007
1(12
1 I)
'
0217
(466
5)
(133
8
I7)2
(
(1)7
6((
5001
'
0391
6038
)
0059
(974
)
9) 3
)0(4
976)
(((1
2)(0
046
)'
-022
4--
025)
(.83
4(I
565
j'
-0,5
901-
10(9
)'
095)
47)4
)
0687
0608
)'
-003
2--
0 10
6)'
0503
(2 0
25
0615
(1 .3
22k'
--0)
4403
39'
0665
I) 1
441'
0170
2
(013
)1'
0235
094)
j'
1041
70)3
)
((66
11
(0 3
95)'
--00
91I -
(308
707
0()'
- 0.
109
- 1.
8511
'
0049
1 33
3)
0041
0301
i'
--0
047
. 3 I
17)
1)00
1(0
04)'
00))
)(0
230
)'
--((
050
-105
6)'
1)9(
33
1041
3)'
0037
(-0
906'
00(
(1 0I)4
oII (1
)9(3
.003
!)'
035)
1) 6
)0
U32
5
((16
1)5(
5
((15
5
03))
((43
8
1) (,4o
s
(I 8
)5
0 34
4
r
valu
es n
oti-s
lgni
ticun
t at 5
per
citn
I(a
) V
iItic
s no
t com
pute
d be
caus
e nu
mbe
r of
obs
erva
tions
was
list
or
Iesv
,
3560
-086
61-
I 355
)72
9910
11)
0637
Io.9
54r
0715
0.45
7I
I05
.907
0.55
9)'
0,18
7
o.ss
or-0
021.
12%
0.41
2)'
0095
.227
)'1.
490
.461
)
0006
099
03l
3993
a)a)
(a)
(10
a)a)
(a)
ci-2
030
- ),
677)
0 02
(I 7
12)'
0.53
0I 1
.209
1'0,
019
0 5)
4)'
110
37)0
a)Ia
)Ia
)al
-0.1
93-0
.87K
v-
2.00
2-
.020
)'-0
.270
1 -
I .2)
6)'
02(8
- I 7
50-.
7384
'0
15
:155
4'
0.4'
I.2
79'
-000
8-.
0.24
09'
0 56
3
311)
301
8302
17)'
0)10
4033
0)'
0095
(0,1
511)
'
0)20
--0.
009
I -00
28)'
- 3.
825
4- 0
.9)6
)'-0
426
--43
.980
)'-0
078
-6.0
69I -
4331
,0
754
4464
30)
-0.1
11-0
3434
'-0
041
-0.9
89)'
1)35
6
36)9
-0,2
93(-
0.35
))'
2365
0 56
6)'
0.19
10,
508)
'-0
.293
-0.0
47I -
0.04
4)'
.830
40.2
9))'
0121
10.2
1 II'
0500
-296
I - I
llor
(4.3
21,
5.02
7)I
III(3
0191
0064
)I 2
791'
0404
,
3029
0076
-270
2-0
,329
0.09
30.
457
1143
0.05
41)
656
-304
9C
224
0007
0013
)'--
0.4)
5)1-
) II)
))'
4467
10.1
141'
I- 1
.167
)'4-
1.34
5)'
42.3
29)
4055
9)40
290)
'-2
6091
3505
)
3(43
-058
0--
0907
--01
190.
564
0.07
1.8
750.
097
077%
-369
81)
553
0.51
3(4
.007
(00)
0)'
(151
4
I-I 1
30)'
(-05
74)'
(-0.
936)
'03
01)'
11.2
03)'
0552
4'-2
4424
5722
)11
194)
'
The distribution of the sample by establishment size groupings is shown in TableA-3. Nevertheless, in spite of the large number of eliminated observations, thesample comprises over 30 percent of thc total nuinbei of establishments for thetwo smallest size groupings and over 70 percent of the total number of establish-ments for the two largest sizegroupings (50-99 and 100 and more people employed)
Catholic tinh'ers it v, Chile
634
I