R&D efficiency and economic performance:危险废物转移联单管理办法
A cross-country analysis using the
stochastic frontier approach
Eric C.Wang∗
Department of Economics,National Chung Cheng University,Minhsiung,Chiayi(621),Taiwan,ROC Received1May2006;received in revid form1September2006;accepted15December2006
Available online30January2007
Abstract
This paper constructs a cross-country production model for evaluating the relative efficiency of aggregate R&D activities.Stochastic frontier methods incorporating translog specification are applied to data of30 countries in recent years.R&D capital stock and manpower are considered as inputs;patents and academic publications are regarded as outputs.Environmental factors that influence R&D performance are also taken into account.The means of efficiency scores without takin
g the environmental effects into account are about0.65in the cross-country study.After controlling for the operating environment,the means increa to about0.85.R&D performance indices show a positive correlation with income level.Policy implications on resources allocation and R&D strategies are discusd.
©2007Society for Policy Modeling.Published by Elvier Inc.All rights rerved.
JEL classification:O32;C24;O50
Keywords:R&D;Efficiency;Stochastic frontier analysis;Operating environment
1.Introduction
The eminent role of scientific R&D in the cour of economic growth has long been empha-sized by governments,enterpris,and scholars.Most contemporary economists have attributed the sustained growth in developed nations to their intensive R&D activities.R&D activity is a well-organized process of knowledge creation,production,diffusion,and application.It involves
∗Tel.:+88652720411x34100;fax:+88652720816.
E-mail address:ecdecw@ccu.edu.tw.
0161-8938/$–e front matter©2007Society for Policy Modeling.Published by Elvier Inc.All rights rerved.
doi:10.1016/j.jpolmod.2006.12.005
346 E.C.Wang/Journal of Policy Modeling29(2007)345–360
innovation in scientific technology,in management measures,and in social and political systems.
A huge volume of resources has been devoted to R&D activities in many countries in recent years. For example,in the year2003,the ratios of gross domestic expenditure on R&D to GDP in the US,Japan,and the EU-25were2.67,3.12,and1.86%,respectively.
Since R&D is one of the most crucial elements in promoting growth,it is argued that any country that us R&D resources inefficiently may be subjected to a growth penalty in the form of a much smaller benefit from R&D investment.If R&D resources are not ud effectively, additional investment may be of little help in stimulating economic growth.However,the existing literature has focud primarily on efforts to engage in new investment and comparatively little attention has been paid to the effective u of R&D resources once they are in place.This is a potentially important omission,since t
he very conditions responsible for economic backwardness may operate through the poor management of the means of engaging in R&D.Therefore,knowing the nature of R&D performance by examining its relative efficiency across countries is thefirst required step for designing policies that intend to improve resource allocation.
The purpo of this paper is to t up an analytical model to evaluate the technical efficiency of aggregate R&D activities in order to offer a comparative scheme to the sampled countries so as to improve the allocation of R&D resources.It also eks to explore the exogenous factors that could affect the countries’respective performances.Our main intention is to emphasize the importance of efficiency in using R&D resources.Following Pakes and Griliches(1984)and Griliches(1990),this study considers R&D as a production process and regards each country as a decision-making unit that conducts R&D.On tting up an inter-country R&D production framework and estimating the performance of the different countries by means of the stochastic frontier production approach,this paper propos a fresh notion as well as a new direction of inquiry into the R&D efforts.The countries making up our sample compri23OECD mem-bers and7non-OECD countries that are intensively engaged in R&D activities.The connection between empirical R&D efficiency and economic growth as well as policy considerations will be discusd at the end of the paper.It is belie
传染病的传播途径ved that the novel approach prented in this paper will contribute to the already substantial body of R&D literature concerned with knowledge-bad economies.
A voluminous literature has already been devoted to investigating the economic aspects and effects of R&D investment.It has been considered that R&D could result in better production technology and also rai the productivity as well as the rates of return on investment at both thefirm and industry levels.Griliches(1986,1990),Mansfield(1988),Goto and Suzuki(1989), Meliciani(2000),Timmer(2003),and Gonzalez and Gascon(2004)have provided theoretical arguments as well as empirical evidence from various industries in many countries.The positive effects of R&D investment on productivity as well as on rates of return are clearly identified. In addition,there are many other issues related to R&D,such as patenting,patent quality and business strategies that have been discusd in the economic literature.Griliches(1990),Ginarte and Park(1997)and Penin(2005)examined the economic aspects of patents.Patent quality and examination procedures were also discusd in King(2003).The relationship between the protection of patents and product standardization strategies was explored by Blind and Thumm (2004).
More and more efforts have been devoted to studying the issues related to academic rearch in universities.Feller(1990)discusd the economic aspects of commercializing basic university rearc
h.Stephan(1996)examined various issues to do with the economics of science and brought the topic of efficiency into consideration.Geuna(1999)surveyed the economic role of university rearch as well as its funding and structure.Adams and Griliches(2000)also examined the
E.C.Wang/Journal of Policy Modeling29(2007)345–360347 rearch performance of the US universities,in which R&D funds and total citations in relation to SCI papers were ud as input and output,respectively.自食其果的意思
Two major approaches to the systematic measurement of production efficiency have been developed in the economics literature.One is data envelopment analysis(DEA),which involves the application of the linear programming technique to trace the efficiency frontier,while the other is the stochastic frontier analysis(SFA)approach,which applies the econometric technique to esti-mating various production/cost frontiers.Aigner,Lovell,and Schmidt(1977),Batte and Coelli (1992,1995),Coelli(1995),Kumbhakar and Lovell(2000),and Ahn,Lee,and Schmidt(2001) contributed to the theoretical development and empirical application of SFA at both thefirm and industry levels.Both SFA and DEA have been widely applied to a variety of issues in economics and the managerial sciences.More recently,Fried,Schmidt,and Yaisawarng(1999)and Fried, Lovell,Schmidt,and Yaisawarng(2002)have propod various methods,which involve the u of DEA,白话翻译
SFA,and the Tobit model,for the purpo of screening out the external effects and statistical noi from the producer’s performance and achieving a more accurate efficiency measure.
The remainder of the paper is organized as follows.Section Two discuss the cross-country R&D production framework and stochastic frontier models.Section Three is devoted to data description and the efficiency test results.Section Four explores the correlation between R&D activities and economic performance.Thefinal ction contains a summary and policy implica-tions.
2.Methodologies and the R&D production function model
This paper ts up a cross-country production framework for R&D activities that is bad on production theory.Each country is regarded as a decision-making unit(DMU)that employs R&D manpower and physical resources as inputs to produce direct tangible outputs,such as patents and scientific publications.1A stochastic frontier approach(SFA)incorporating a translog specification is applied to estimating the relative R&D efficiency of each country.
2.1.The R&D production function
In a way similar to the knowledge production function t up in Pakes and Griliches(1984)and Grilich
es(1990),this paper considers international R&D activity in the context of a cross-country production function.The R&D production function applied to each country is assumed to be well-behaved and to exhibit variable returns to scale.In the cross-country production framework ud in this paper,it is presumed that all countries have the same underlying aggregate production function in terms of standardized quantities of outputs and inputs,but that they may operate on a different part of it.2
The cross-country R&D production function has the following general form:
y kt=f(x kit)k=1,...,K(country),i=1,...,N(inputs)(1) where y kt is the R&D output of country k at time t and x kit is input i of country k at time t.
1An alternative measure of R&D performance might be in terms of GDP or productivity growth.However,it is considered that GDP or productivity is a somewhat indirect measure of R&D efforts.There are many other factors, besides R&D,that can contribute to GDP or productivity growth.It is preferable to adopt a direct measure of R&D output in this paper.
2This assumption is similar to that adopted by Lau and Yotopoulos(1989)and many subquent studies regarding multinational economic growth and productivity.
348 E.C.Wang/Journal of Policy Modeling29(2007)345–360
水果的英语
All inputs and outputs in this function are assumed to be homogeneous and there is no budget constraint of any form.
The wide ranges of variation in inputs resulting from the u of pooled inter-country data necessitate the u of theflexible functional form f(.)above.Aflexible functional form also allows for the possibility of non-neutral returns to scale and technical change.A commonly ud flexible functional form of production is the translog form,which has the property of a varying elasticity of substitution between inputs.Furthermore,it is a general form of many popularly ud production functions,such as the Cobb-Douglas form.A translog production function at time t can be written as:
ln y=b0+Σb i(ln x i)+Σb ii(ln x i)2+ΣΣb ij(ln x i)(ln x j)(2) where y reprents the output volume,and x i and x j are the inputs i and j,respectively.ln is the symbol for a natural logarithm.The parameters b i,b ii,and b ij to be estimated are independent of any particular individual country and supply a common link among the aggregate R&D production functions of the different countries.This provides a basis for testing the maintained hypothesis of the cross-country function approach,namely,that there is a single aggregate R&D production function for all the countries.3
2.2.The stochastic frontier approach for R&D efficiency
The aim of this paper is to estimate a cross-country production possibility frontier and to asss the technical efficiency of each country in relation to that frontier bad on the aggregate R&D efforts.Batte and Coelli(1992)have propod a time-varying model for evaluating the technical efficiency that takes the form of a stochastic frontier production function for panel data,and which hasfirm(or DMU)effects that are assumed to be distributed as truncated normal random variables. The model may be expresd as:
y kt=f(x kt)+(V kt−U kt)k=1,...,K(country),t=1,...,T(time)(3)
where y kt is the logarithm of production of the k th country in the t th time period;x kt is a vector of(transformation of the)input quantities of the k th country in the t th time period;and f(x kt)is a flexible function,such as the translog specified above.The V kt are random variables,which are assumed to be iid N(0,σ2V)and independent of the
U kt={exp[−η(t−T)]}U k(4)
where the U kt are non-negative random variables,which are assumed to account for technical inefficiency in production and are assumed to be iid as truncations at zero of the N(μ,σ2U).ηis an unknown scalar parameter to be estimated.In addition,the panel data need not be complete (i.e.they
may be unbalanced panel data).under反义词
In the specification of equation(4),if the k th DMU is obrved in the last period of the panel, period T,then U kT=U k,becau the exponential function,exp[−η(t−T)],has a value of one when t=T.Thus the random variable,U k,can be considered to be the technical inefficiency effect for the k th DMU in the last period of the panel.For earlier periods in the panel,the technical efficiency effects are the product of the technical inefficiency effect for the k th DMU in the last period of 3For the properties and restrictions of the translog production function,e Christenn,Jorgenson,and Lau(1973).
E.C.Wang/Journal of Policy Modeling29(2007)345–360349 the panel and the value of the exponential function,exp[−η(t−T)],who value depends on the parameter,η,and the number of periods before the last period of the panel,−(t−T)≡(T−t).If the parameterηis positive,then−η(t−T)≡η(T−t)is non-negative and so exp[−η(t−T)]is not smaller than one,which implies that U kT≥U k.Converly,ifηis negative,then−η(t−T)≤0
and so U kT≤U k.
For the purpo of estimation,the parametersσ2V andσ2U are replaced withσ2=σ2V+σ2U andγ=σ2U/(σ2V+σ2U),which are computed from the maximum likelihood(ML)estimates.The paramet
er,γ,must lie between0and1and thus this range can be arched to provide a good starting value for u in an iterative maximization process.
The ability of a production unit to transform inputs into outputs is affected not only by its tech-nical efficiency but also by exogenous environmental factors.In an effort to analyze the difference in efficiency of various production activities,a number of empirical Pitt and Lee, 1981)estimated stochastic frontiers and predictedfirm-level efficiencies using various functions, and then regresd the predicted efficiencies onfirm-specific variables in an attempt to identify some of the reasons for differences in predicted efficiencies betweenfirms in industries.How-ever,this two-stage estimation method suffers from a problem of inconsistency in its assumptions regarding the independence of the inefficiency effects in the two estimation stages.4Batte and Coelli(1995)propod a technical efficiency frontier model,which accounts for factors that may influence the efficiency of a production unit and permits the u of panel data.
Batte and Coelli’s(1995)frontier model may be specified as:
淄博美食
y kt=f(x kt)+(V kt−U kt)k=1,...,K(country),t=1,...,T(time)(5) where y kt,x kt,and f(x kt)are defined as in Batte and Coelli’s(1992)model.The V kt are random variables,which are assumed to be iid N(0,σ2
V))and independent of the U kt.The U kt are non-negative random variables,which are assumed to account for technical inefficiency in production and are assumed to be iid as truncations at zero of the N(m kt,σ2U)distribution.The m kt in the N(.) distribution is specified as
m kt=z ktδ,(6) where z kt is a vector of environmental variables,which may affect the efficiency of a production unit andδis a vector of parameters to be estimated.The parametersσ2V andσ2U are once again replaced withσ2=σ2V+σ2U andγ=σ2U/(σ2V+σ2U)for the purpo of estimation using the ML method.
3.Data and empirical results
3.1.Data sources and definition of variables
A total of30countries form the sample in this study.Among them,23are OECD members and 7are non-OECD economies that engage in R&D intensively.Official data on the R&D inputs and outputs of the countries compiled and relead by international organizations,governments, and private institutions are collected.Detailed data sources and sample countries are listed in the Appendix.
草鱼的做法大全
4See Kumbhakar,Ghosh,and McGuckin(1991).
