Study with the several resources on Docsity
Earn points by helping other students or get them with a premium plan
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Community
Ask the community for help and clear up your study doubts
Discover the best universities in your country according to Docsity users
Free resources
Download our free guides on studying techniques, anxiety management strategies, and thesis advice from Docsity tutors
This paper investigates the relationship between capital and labor employed in research and its output, using a panel of 31 countries over nine years. The study finds that investing in researchers contributes more to research outcomes than investing in research capital, especially due to the large number of researchers in higher education. The paper also discusses the importance of research and development (R&D) in economic growth and productivity.
What you will learn
Typology: Study notes
1 / 21
This page cannot be seen from the preview
Don't miss anything!
This paper can be found at the College of Business and Economics Working Paper
Series homepage:
http://business.wvu.edu/graduate-degrees/phd-economics/working-papers
Scienti c output: labor or capital intensive? An
analysis for selected countries.
Abstract Scienti c research contributes to sustainable economic growth environments. Hence, policy-makers should understand how the dierent inputs namely labor and capital are related to a country's scienti c output. This paper addresses this issue by estimating output elasticities for labor and capital using a panel of 31 countries in nine years. Due to the nature of scienti c output, we also use spatial econometric models to take into account the spillover eects from knowledge produced as well as labor and capital. The results show that capital elasticity is closer to the labor elasticity. The results suggest a decreasing return to scale production of scienti c output. The spatial model points to negative spillovers from capital expenditure and no spillovers from labor or the scienti c output.
Keywords: Scienti c output; capital; labor; spillover eects JEL Classi cation: O32, F01, O
∗This paper was the co-recipient of the Best Paper by Graduate Student in Economics Award at
the 54th Annual Conference of the Academy of Economics and Finance - Charleston SC, 2017. †Regional Research Institute, School of Natural Resources. e-mail: elhamerfanian@mix.wvu.edu ‡W. Marston and Katharine B. Becker Fellow. Department of Economics, Regional Research
Institute. e-mail: amneto@mix.wvu.edu
econometric approach; section 4 presents the data used; section 5 discusses the results; and section 6 concludes.
2 Research and Development (R&D) and Economy
Growth
Schumpeter (1942) and Solow (1957) are pioneers in the study of innovation and technical changes as engines to production and economic growth. King (2004) argues that the ability to judge a nation's scienti c standing is vital for both governments and society, as the result of scienti c eorts may be seen in higher economic growth rates and more economic outputs. These eventually re ect the increase in social welfare. Long-term economic growth needs a sustainable fuel, which could be provided by innovations. In long run, the ability of a nation to improve the standard of living passes through increasing the output-to-input ratio. A broad overview in productivity triggers is presented in the literature, most of them emphasizing technology and research. For instance, Guellec and van Pottelsberghe De La Potterie. (2001) analyzes 16 OECD countries and nds that R&D is an important factor for productivity and economic growth. Rouvinen (2002) studies four issues in R&D and productivity. His results suggest that R&D investment in uences productivity not vice versa. Bravo-Ortega and Marin (2010) provide evidence that corroborates Rouvinen (2002)'s results. More recently, Eid (2012), using country-level data for 17 high-income OECD countries, measures the correlation between R&D and productivity growth and nds there is a lag between them. In the tradition of the knowledge-capital model of Griliches, Doraszelski and Jaumandreu (2013) develop a model to investigate the correlation between R&D and productivity. The authors nd R&D expenditure has a key role in productivity across rms. For scientometrics, scienti c publication is the engine of economic growth. There- fore, the knowledge spillover discussion becomes relatively more important. This dis- cussion started Alfred Marshall (Carlino et. al, 2001), and it still gets the attention of many economists. Some of the researches on knowledge spillovers are summarized below. Griliches (1986) nds that after controlling for industry-speci c xed eects, the eects of research on productivity growth is cut by about fty percent. The author explains this is because of spillovers within the industry. Jae (1989) and Jae et al. (1993) show that spillovers are industry and geographically localized. Varga (2000) applies the Griliches-Jae knowledge production function and expands it to a hierar- chical version to test the knowledge spillovers in U.S. metropolitan areas, nding that research universities can increase the regional production. More recently, Elhorst and Zigova (2014) nd no evidence of cross-fertilization eects across nearby universities, which corroborates the Bonaccorsi and Daraio's (2005) results. However, the authors argue that collaboration has a positive eect on research productivity.
In an oversimpli cation scenario, research and development (R&D) has two major inputs and two major outputs, capital and labor for the former, and patents and publications for the latter. In this work, we focus only on the publication output. McAllister and Wagner (1981) examine the relationship between R&D expenditure and the number of papers published in a sample of 500 universities and colleges in the United States. For each of 11 elds of science that the authors consider, there is a strong positive relationship between R&D expenditure and the number of pub- lications. Focusing only on late industrial countries, Amsden and Mourshed. (1997) examine the scienti c publication, patent and technological capabilities. While the authors expect a high growth rate of GDP and scienti c publications to be posi- tively correlated, they nd the high correlation in countries like South Korea, China and Singapore rather than in countries such as Turkey, Argentina, Brazil, Chile and Mexico. Shelton (2008) compares American and European publications and nds that the eectiveness of research investment is more signi cant than the number of scientists for scienti c outcomes. Sharma and Thomas. (2008) nds that the number of e - cient countries in the R&D sector varies based on the assumption the authors made. Crespi and Geuna. (2008) nd a strongly positive long-run relationship between R&D expenditure and the number of publications with an optimum lag of 6 years. Adams et al. (2013) look at Brazil, Russia, India, China and South Korea, known together as the BRICK nations. They nd Brazil stands out as dierent from the others. While a natural knowledge economy is a leading area in Brazil, research policy, physics, chemistry, engineering and materials are the leading areas in Russia, India, China, and South Korea. Akhmat (2014) the relationship between educational indicators and research out- comes in the top twenty countries. The results indicate that education expenditures and the number of publications have a one-to-one relationship. In a series of papers, Meo and Usmani (2014) and Meo et al. (2013b, 2013a found among Asian countries, Middle East, and European countries a positive correlation between spending on R&D and the number of research publications, while in all the sub-samples the results show no correlation between GDP per capita and the total number of publications. They also conclude that the research outcome depends on the ratio of R&D expenditure to the total GDP not the absolute R&D expenditure.
3 Model and Econometric Approach
Assume countries produce scienti c research following a Cobb-Douglas production function in which there are two main inputs: capital and labor.
Y = KαLβ^ (1) By assuming a Cobb-Douglas, we implicitly assume that there is no heterogeneity between countries. At rst this may seem unreasonable; however, given the easy
The World Bank provides the number of scienti c and technical journal articles for all the countries around the world. The dependent variable for all the speci cations is based on the World Bank. Explanatory variables for scienti c inputs in the World Bank model includes the number of researchers in R&D and gross capital formation. This dataset includes 31 countries in a panel of nine years from 2003 to 2011. The OECD explanatory variables include the full-time equivalent researchers in total, and we further break it down to business enterprise, government, and higher education sectors. The total labor cost and the total capital expenditure in research are the capital related input in OECD countries analysis This dataset contains 22 countries from 2003 to 2011. The scienti c output information is available for 46 countries from 1996 to 2011 in an unbalanced set up. Because we believe the use of spatial econometrics techniques are very important in this study, we created a balanced panel of countries that max- imized the number of observations. Moreover, we chose to use of both World Bank and OECD datasets because one may argue that gross formation of capital is not the best measure for R&D capital investment and the measure for researchers should be disaggregated. Therefore, we attempt to deal with these possible concerns, but to have a balanced panel, we have to drop 9 other countries that were in the World Bank sample. Table 1 provides the summary statistics for the data.
<INSERT TABLE 1 >
5 Results
The results are divided into two parts. First we present the results without any spatial spillovers and then we introduce such results. As explained in the previous sections, we believe that the spatial spillovers are important both theoretically and empirically. Tables 2 to 5 present the results with no spatial dependence. The analysis will focus on model (4), our preferred speci cation. Table 2 presents the results using the number of researchers provided by the World Bank (WB) and the gross capital formation, also provided by the World Bank. The results show no in uence of capital on the scienti c output, while the elasticity of labor is positive and statistically signi cant. To further investigate^3 this relationship, we look at another data source. Tables 3 presents results using labor cost as a proxy for the number of researchers in R&D and capital cost, Table 4 uses the total number of researchers, and Table 5 disaggregates the researchers into three categories: business enterprise, government and higher education. The dependent variable remains the number of articles produced reported by the World Bank. The results for the preferred model (4) from tables 3 to 5 show a positive and statistically signi cant result for both capital and labor. According to the elasticity values, there is decreasing returns to scale relation, as the sum of both elasticities
(^3) Another robustness check performed was the analysis for unbalanced panels in all scenarios discussed. The results remain similar in terms of sign and signi cance of the estimated coe cients. These results are available upon request.
are less than one in every case. It is interesting to note when using the number of researchers instead of the labor cost (tables 4 and 5) the results suggest the capital and labor elasticities have similar magnitude.
<INSERT TABLES 2 TO 5>
As discussed in the previous sections, it is important to consider the spillover eects of knowledge both theoretically and empirically. Therefore, we present in tables 6 to 9 the spatial results for the regressions presented in tables 2 to 5. We present four spatial models: SAR, SDM, SDEM and SLX; however, we will focus the analysis on the SDM model as we believe it is the best model because it considers spillovers from the dependent variables (articles) and the explanatory variables (inputs). In terms of the weight matrix, we used the k-nearest neighbors weight matrix with k equals to 1. This was the weight matrix that captured the most spatial dependence. For the World Bank sample (table 6) we observe that the results remain similar to those of table 2, but there is an extra weight on the labor elasticity. Also, there is no evidence of articles or input spillover, which suggests countries have access to the same information regardless if they are neighbors. As for the OECD sample (tables 7 to 9), there is statistically signi cant negative spillover of capital expenditure on R&D. This suggests that investing in R&D has a negative eect on knowledge output in close-by countries. There is no spillover of labor inputs nor of scienti c outputs. Also, the countries own labor and capital inputs have positive and statistically signi cant results. One possible concern is the use of the geographical matrix to do the spatial anal- ysis. We would argue that this matrix is good for several reasons. Firstly, we need the weight matrix to be exogenous to our estimation, and the geographic matrix ts this requirement. Secondly, it is well established in the literature that distance has an inverse relation to economic outcomes. Lastly, several authors (Jae, 1989; Jae el al., 1993; Varga, 2000) show that geographical proximity is important for spillovers.
<INSERT TABLES 6 TO 9>
6 Conclusions and Implications
The objective of this paper is to understand the production of scienti c output for several countries. More speci cally, we wanted to investigate the relation of capital and labor employed in research to its output. We used a balanced panel of 31 countries and 9 years to estimate the capital and labor elasticities and then employed spatial models in order to capture possible spillovers. The main results can be divided into two: rstly, capital and labor seems to have similar importance in terms of producing scienti c output; and when disaggregated, researchers in the business enterprise have zero output elasticity. Secondly, in terms of spillovers, there seems to be a negative
Griliches, Z. (1986). Productivity, r&d, and basic research at the rm level in the 1970s. American Economic Review, 76(1):141154.
Guellec, D. and van Pottelsberghe De La Potterie., B. (2001). R&d and productivity growth: panel data analysis of 16 oecd countries. Technical report, OECD.
Jae, A. B. (1989). Real eects of academic research. American Economic Review, 79(5):957970.
Jae, A. B., Trajtenberg, M., and Henderson., R. (1993). Geographic localization of knowledge spillovers as evidenced by patent citations. The Quarterly journal of Economics, 108(3):577589.
King, D. A. (2004). The scienti c impact of nations. Nature, 430(7):311316.
McAllister, P. R. and Wagner, D. (1981). Relationship between r&d expenditures and publication output for us colleges and universities. Research in Higher Education, 15(1):330.
Meo, S., MAsri, A. A., Usmani, A., Memon, A., and Zaidi, S. (2013a). Impact of gdp, spending on r&d, number of universities and scienti c journals on research publications among asian countries. PLOS One, 8(6):18.
Meo, S. and Usmani, A. (2014). Impact of r&d expenditures on research publications, patents and high-tech exports among european countries. European Review for Medical and Pharmacological Sciences, 18(1):19.
Meo, S., Usmani, A., Vohra, M., and Bukhari, I. (2013b). Impact of gdp, spending on r&d, number of universities and scienti c journals on research publications in pharmacological sciences in middle east. European Review for Medical and Phar- macological Sciences, 17(20):26972705.
Rosenberg, N. (2014). Innovation and economic growth. Technical report, OECD.
Rouvinen, P. (2002). R&d-productivity dynamics: Causality, lags, and ?dry holes?. Journal of Applied Economics, 5(1):23156.
Schumpeter, J. A. (1942). ICapitalism, socialism and democracy. London: Unwin Paperbacks.
Sharma, S. and Thomas., V. (2008). Inter-country r&d e ciency analysis: An appli- cation of data envelopment analysis. Scientometrics, 76(3):483501.
Shelton, R. D. (2008). Relations between national research investment and publication output: Application to an american paradox. Scientometrics, 74(2):191205.
Solow, R. M. (1957). Technical change and the aggregate production function. The Review of Economics and Statistics, 39(3):312320.
Varga, A. (2000). Local academic knowledge spillovers and the concentration of economic activity. Journal of Regional Science, 4(2):289309.
Tables
Table 2: World Bank Sample Results
Dependent variable:
Articles
Fixed Eects
Gross Capital Formation
∗^
Total
of Researchers WB
∗∗∗
∗∗∗
∗∗∗
∗∗∗
Constant
∗∗∗
Year FE
Country FE
Observations
2
Adjusted R
2
Note:
∗ p<
∗∗
p<
∗∗∗
p
0.01 In parenthesis, we present the standard deviation.
The variables in the regression are in log form, so the coe cients can be interpreted as theoutput elasticities of capital and labor.For the xed eects model we used the package lfe in R
Table 3: OECD Sample Results
Dependent variable:
Articles
Fixed Eects
Capital Expenditure on R&D
∗∗∗
∗∗∗
∗∗∗
∗∗∗
Labor cost on R&D
∗∗∗
∗∗∗
∗∗∗
∗
Constant
∗∗∗
Year FE
Country FE
Observations
2
Adjusted R
2
Note:
∗ p<
∗∗
p<
∗∗∗
p
0.01 In parenthesis, we present the standard deviation.
The variables in the regression are in log form, so the coe cients can be interpreted as theoutput elasticities of capital and labor.For the xed eects model we used the package lfe in R
Table 5: OECD Results, disaggregate researchers
Dependent variable:
l.articles
Fixed Eects
Capital Expenditure on R&D
∗∗∗
∗∗∗
∗∗∗
∗∗∗
Researchers in Business Enterprises
∗∗∗
∗∗∗
Researchers in Government
∗∗
∗∗∗
∗∗
∗∗
Researchers in Higher Education
∗∗∗
∗∗∗
∗∗∗
∗∗∗
Constant
∗∗∗
Year FE
Country FE
Observations
2
Adjusted R
2
Note:
∗ p<
∗∗
p
∗∗∗
p
0.01 In parenthesis, we present the standard deviation.
The variables in the regression are in log form, so the coe cients can be interpreted as theoutput elasticities of capital and labor.For the xed eects model we used the package lfe in R
Table 6: World Bank Sample Spatial Results
Dependent variable:
Article
Gross Capital Formation
Total
of Researchers WB
W*Articles
W*Gross Capital Formation
W*Total
of Researchers WB
W*Error Term
Year FE
Country FE
R squard
Number of observations
Note:
∗ p
∗∗
p<
∗∗∗
p<
0.01 In parenthesis, we present the standard deviation.
The variables in the regression are in log form, so the coe cients can be interpreted as theoutput elasticities of capital and labor. For the spatial models SAR, SDM and SDEM we usedthe package splm in R,and for the SLX the package lfe.
Table 8: OECD Expenditures Spatial Results
Dependent variable:
l.articles
Capital Expenditure on R&D
Total
of Researchers OECD
W*Articles
W* Capital Expenditure on R&D
W* Total
of Researchers OECD
W*Error Term
Year FE
Country FE
R squard
Number of observations
Note:
∗ p
∗∗
p<
∗∗∗
p<
0.01 In parenthesis, we present the standard deviation.
The variables in the regression are in log form, so the coe cients can be interpreted as theoutput elasticities of capital and labor. For the spatial models SAR, SDM and SDEM we usedthe package splm in R,and for the SLX the package lfe.
Table 9: OECD Sample Spatial Results, Disaggregated Researchers
Dependent variable:
l.articles
Capital Expenditure on R&D
Researchers in Business Enterprises
Researchers in Government
Researchers in Higher Education
W*Articles
W*Capital Expenditure on R&D
W* Researchers in Business Enterprises
W* Researchers in Government
W* Researchers in Higher Education
W*Error Term
Year FE
Country FE
R squard
Number of observations
Note:
∗ p
∗∗
p<
∗∗∗
p<
0.01 In parenthesis, we present the standard deviation.
The variables in the regression are in log form, so the coe cients can be interpreted as theoutput elasticities of capital and labor. For the spatial models SAR, SDM and SDEM we usedthe package splm in R,and for the SLX the package lfe.
