Research funding and research production: A cross country analysis


Seminar Paper, 2006

33 Pages


Excerpt

Contents

1 Introduction

2 The productivity of UK universities

3 A cross country analysis
3.1 Data sources and method
3.2 Publications 2002
3.3 Publications 2003
3.4 Analysis
3.4.1 Learning-by-Doing
3.4.2 Path dependence
3.5 Conclusion

4 Critical evaluation

5 Possible extensions

6 Summary

7 References

8 Appendix

1 Introduction

Productivity can refer to different scientific fields.1 In economics, productivity is the amount of output created per unit input used. In corporate finance productivity is defined by the current year's sales to expense ratio over the previous year's sales to expense ratio. In linguistics, the degree to which a grammatical process can be extended to new cases is meant, whereas Social Productivity is the ratio of the volume of a group’s collaborative output produced per unit input (e.g. time) used.

In any case an increase of productivity2 is useful. In economics an increase of productivity reduces the costs or rises the benefits, respectively. Therefore it is interesting to examine the existing differences in order to improve the productivity.

The paper by Crespi and Geuna headlined „The productivity of UK universities“ also deals with this topic.3 They analysed the determinants of the three most common university research outputs, among others publications as well. For these publications for example they estimated significantly different lag structures, which means that the science system does not respond uniformly to changes in funds. I’ll expand on their work and their final results in chapter 2.

Research funding as well as research production differ among the different European countries. Furthermore the productivity regarding the research activity differs, too. In chapter 3 I’ll compare some European countries regarding their research activity and their efficiency. The idea is to find out which factors cause the differences in their productivity in order to describe how countries with a lower productivity could improve their performance. Using the R&D expenditure as input factor and the scientific publication as output I’ll examine if Learning-by- Doing (LbD) or path dependence play a crucial role concerning this question.

Furthermore I’ll try to find out whether some other factors that influence the relation between input and output or if time lags can be found.

At first sight LbD doesn’t seem to be a satisfying solution, as short run changes are impossible. But at least it shows, that the difference doesn’t result from any mistakes. Instead it reveals out that it will simply take some time to improve productivity. Path dependency as a solution would mean, one could check if it can be reversed by calculating the switching costs. But in fact, path dependent processes are often irreversible.

Chapter 4 includes a critical evaluation regarding the data sets used, the approach itself and assumptions I made.

Chapter 5 points out some other possible methods and I’ll refer to some other input or output factors, such as patents and researchers.

Finally the fundamental results of this work will be summarised in chapter 6.

2 The productivity of UK universities

The paper by Crespi and Geuna headlined „The productivity of UK universities“ deals with the question how increased funding for research in the UK can lead to increasing levels of scientific output and therefore to better economic performance.4

They analysed the determinants of several research outputs. For a measure of new knowledge they use publications and citations and highly qualified human ressources have been proxied by the number of graduate students that have completed their studies. They used a dataset that includes information for the 52 old UK universities across 29 scientific fields for a period of 18 years.

As information about publications and citations is only available at field level they had to aggregate the scientific fields into 4 broad categories. These macro fields are natural sciences, engineering, medical sciences and social sciences. The authors developed the standard production function model of Adams and Griliches (1996) and they also use a Polynomal Distributed Lag (PDL) Model to search for a lag structure. With these data sources and models they estimate the science production function for the macro fields and examine the changes in productivity across fields. They also use control variables. Using information about the number of undergraduate students by field and year they control for the impact of teaching intensity on research output. Furthermore they take into account that the research output can be affected by factors specific to the university.

For the output measured by publications they find out that the lag structures are significantly different across fields. Social sciences have a relatively important impact in the short run but the effects diminish over time. The situation in the natural and medical science is the reverse, while in the case of engineering a parabolic function suggests that the impact is most important towards the middle of the time period. This is a very important result as it points out to a differential impact of a given increase in spendings over time.

For the examination of citations they got similar results as for publications. The citation output tends to respond more quickly to an increase in investment in social science than in the other fields. Medical science shows its largest research impact only at the end of the time period, while the natural science and engineering are in between.

Looking at graduate students as a science output Crespi and Geuna found out that the patterns appear to be quite different from those for publications and citations. As suggested by the authors the reason for this could be that graduate students are a research ouptut of a completely different nature to publications and citations. In the case of graduate students medical science, engineering and natural science show the strongest impact very quickly, while the impact of social science appears not until the end of the time period.

For publications and citations they estimated significantly different lag structures with a long lag for medical sciences before full returns of an increase in spending was achieved, but social sciences seeing results in the first five years. Furthermore the authors obtained statistically significant and important coefficients for some of the control variables. The undergrate teaching variable was negative in natural science, medical science and engineering. This coefficient was always negative, suggesting that large undergraduate teaching loads have a disruptive effect on scientific production. In social sciences a positive impact of undergraduate teaching towards scientific production can be found.

Crespi and Geuna also focused on the efficiency with which the stock of knowledge is applied in order to create the different research outputs. They computed field specific total factor productivities (TFPs) that capture the evolution of the scientific opportunities in each field as well as the effects of changes in organisational practice, resource allocation and management. The result is an upward trend in the productivity indices in all macro fields, suggesting that there is a clear improvement in the efficiency and technological opportunities of the system. In all fields and for the three used outputs of scientific research there has been a slowdown in productivity growth rates. Finally the authors point out the fact that their work aimed to test the feasibility of using econometric models to produce results that could contribute to the development of science policy but not to produce exact indicators of the dynamics of the science system on the basis of which strong policy conclusions can be drawn. Rather the inherent shortcomings in the meaurement of the output of the scientific activity and the limitation of the available input data call for extreme caution in the interpretation of our results.

3 A cross country analysis

As mentioned in the introduction research funding as well as research production differ among the different European countries. Furthermore the productivity regarding the research activity differs, too. In this chapter I’ll compare some European countries regarding their research activity and their efficiency thereby. The idea is to find out what causes the differences in their productivity. Thus I could describe how countries with a lower productivity could improve their performance. Using the R&D5 expenditure as input factor and the scientific publication as output I’ll examine if LbD or path dependence play a crucial role concerning this question. Furthermore I’ll try to find out if there some other factors that influence the relation between input and output can be found.

3.1 Data sources and method

To examine the difference in S&T6 performance I used the R&D expenditure per capita as input factor7. For the scientific output I used the scientific Publications per head as they reflect the research capacity and knowledge pool of a country.8

I used Data from the European Commission - mainly from the Key figures 2003/2004 and the Eurostat homepage. The Data sets I used are listed in the Appendix.

The R&D expenditure was given as % of the GDP (Gross Domestic Product).

To get the per capita value I had to convert them. For the GDP I could find two complete records. The first was the GDP per capita in PPS (purchasing power standards), the second was the GDP by current prices.9 The PPP (purchasing power parity) that is used to calculate the PPS can vary with the specific basket of goods used or with the preferences of the respective inhabitants.10 As this work is about Europe only it is likely to assume that the basket of goods as well as the preferences are almost identical, so that using the GDP in PPS would not falsify the results. Figures 1 and 2 show that it makes almost no difference, which of them is used. Another indication for this are the R², that don’t differ very much. The R² for current prices is 0,44 and the one for PPS is 0,4159 which means that both can explain more than 40% of Publication differences.11 The blue line could be seen as some kind of an efficiency curve, for all efficient countries lying on it.

The next questions was, which years to compare. My intuition was, that expenditures have an effect of the output in the same year in which they are spent. Figures 3 and 4 show that the outcome is almost the same, no matter if publications and expenditure of the same years are used or if I take the expenditures of the previous year. The R² is almost the same as well - 0,691 if different years are used and 0,6867 for publications and expenditures of the same year.

For publications only the data sets of the years 2002 and 2003 were available. For the year 2001 I found a data set for 15 member states only, as there haven’t been more member states at that time.

A complete data set for R&D expenditures was available for the years 2000 and 2001. In addition to that I got an average annual growth rate of the R&D expenditure. In order to get the expenditure of the year 2003 I would have had to convert them twice. For I didn’t want to falsify the result I decided to take the expenditure previous year, hence why I had to use the growth rate only once.

3.2 Publications 2002

Figure 4 shows the scientific Publications of 2002 to the expenditure of 2001 of 31 european countries. Of course these are not only member, but also nonmember states, as well as in the following work.

The efficient states are Romania, Latvia, Bulgaria, Estonia, Greece, Spain, Slovenia, the United Kingdom, the Netherlands, Denmark and Switzerland. The black curve is a trend line and gives rise to the idea that there are decreasing returns to scale. This means that an increase in output requires a relative higher increase of the input factor. I’ll refer to that later, as I will try to find out if the reason for the differing efficiency can be explained by Learningby-Doing or path dependence.

3.3 Publications 2003

Figure 5 shows the Publications of 2003 to the expenditures of 2002. The trend line has got the same shape as before. So again the output seems to increase less than the input and therefore the productivity seems to decrease as well. The efficient states are almost the same, as well. The only difference is that the UK isn’t efficient any more while Poland is on the curve.

3.4 Analysis

What happens to the output, if the countries increase or decrease their expenditures. Figure 8 shows the percentage changes of the publications in relation to the changes in expenditures. Most of the countries are lying in the first and third quadrant. In the first quadrant increasing the input leads to an increase of the output and in the third quadrant a decrease of the input decreases the output. As this follows the intuition, it is rather uninteresting. More interesting are those countries lying in the other quadrants. Bulgaria for example raised the input, but the output decreased. A much bigger difference is shown by Turkey and Portugal. They decreased the output, but the number of Publications increased. This shows that there must be something else influencing the scientific output.

3.4.1 Learning-by-Doing

As I suggested in chapter 3.2 LbD might be the reason for the different outcomes of the countries‘ research activity. LbD refers to the capability of workers to improve their productivity by regularly repeating the same type of action. The increased productivity is achieved through practice, self-perfection and minor innovations.12

Figures 6 and 7 show the productivity of the efficient countries.13 In both diagrams Romania is the one with the lowest output, whereas Switzerland is the one with the most publications.

On average the productivity decreases as the output increases. A statement about decreasing returns to scale and decreasing productivity only holds, if all these countries are completely comparable. But these countries are not all completely comparable, as the intuition says and as some outliers show. For a further examination more data is needed. Nevertheless I tried to find out at least a little bit using the data of 3 years (2001, 2002 and 2003) and 4 countries. If there is LbD, there must be an increase in productivity, as well. I examined only one country at a time, trying to find out if there is LbD as an explanation of the differences in productivity.

Figure 9 shows the graph of Finland. From the year 2001 to 2002 there was a high increase in expenditure, but a decrease in publications. From 2002 to 2003 we can see a slight increase of the input but a much higher output then one would expect. In this case there could be a time lag. Skipping the year 2002 we would get a logic correlation between change of input and the resulting change of output such that an increase in input leads to an increase of the output. So the increase of the expenditures in 2001 might lead to a greater number of publiations at a later point of time. Of course the time lag could be larger, such that the increase in output of the year 2003 is due to an increase in inputs that took place before 2001. Therefore the increase in inputs of 2001 would lead to an increase of output only after 2003.

The productivity of Sweden is shown in Figure 10. This graph also shows, that there must be a time lag or something else influencing the productivity. First there is a slight decrease of expenditures that leads to a relatively large decrease in publications. From the year 2002 to the year 2003 we find a very much larger decrease in the input factor but an increase of the output. For this country skipping the year 2002 would also make sense, as from 2001 to 2003 we find a decrease in expenditures and therefore a decrease in publications. Figure 11 shows that the time lag from Germany must be larger or that there must be some other factor that influences the relation between input and output.

First a little decrease of the input leads to a large decrease of the output and then a very high increase of the expenditures leads to a relatively small increase of publications.

The productivity of Luxembourg as shown in Figure 12 is the only one that seems to be reasonable. An increase in expenditure increases the publications in each period. It gives rise to the assumption that there is no LbD, as the productivity decreases over time.14 However, this could be only a coincidence and maybe we could find a time lag or some other factor for this country as well.

Finally I can neither verify nor reject the existence of LbD. To be capable of saying anything about the existence of LbD the time lag structure - if one exists - has to be estimated. In order to do so, a further examination with more than a 3-year-period is necessary and therefore a bigger data set. Furthermore I can not eliminate the thought of some other factors influencing the productivity. I’ll refer to that in chapter 5, listing some other possible explanations, such as other input factors that affect the output.

3.4.2 Path dependence

Path dependence describes the dependence of current states on past decisions.15 Some authors use path dependence to mean simply "history matters", while others use it to mean that institutions are self reinforcing. The outcome of a path dependent process will often not converge towards a unique equilibrium but instead reach one of several equilibria (sometimes known as absorbing states).16

[...]


1 Wikipedia (2006): Productivity

2 efficiency, which is the amount of output produced relative to the amount of resources that go into the Produktion, respectively

3 G. Crespi and A. Geuna (2006)

4 Crespi and Geuna (2006)

5 Research and Development

6 Science and Technology

7 In the following I’ll talk about expenditures and publications only, whereas the per capita values are meant.

8 Key Figures 2002, p.47.

9 The purchasing power standard is a notional unit independent of the national currency that eliminates biases due to differences in the price level. It is calculated by means of purchasing power parities, which are computed by a mix of representative goods and services of a country.

10 Wikipedia (2006): Purchasing power parity

11 R² indicates, which percentage of the deviations from the mean value are explained by the correlation. Example: R²=0,44 means that expenditures can explain 44 percent of Publication differences.

12 Wikipedia (2006): Learning-by-doing

13 measured as publications divided by the expenditure

14 that means that in order to increase the output a higher amount of input is necessary over time.

15 Example: Driving left as in the UK or right as in Germany. Both option are technically equal, but once the the decision to drive on the right (left) side was made, it became permanent because of the huge switching costs a change would imply.

16 Wikipedia (2006): Path dependence

Excerpt out of 33 pages

Details

Title
Research funding and research production: A cross country analysis
College
University of Constance
Course
Seminar "Economics of Science"
Author
Year
2006
Pages
33
Catalog Number
V65517
ISBN (eBook)
9783638580632
File size
1259 KB
Language
English
Tags
Research, Seminar, Economics, Science
Quote paper
Saskia Schierstädt (Author), 2006, Research funding and research production: A cross country analysis, Munich, GRIN Verlag, https://www.grin.com/document/65517

Comments

  • No comments yet.
Read the ebook
Title: Research funding and research production: A cross country analysis



Upload papers

Your term paper / thesis:

- Publication as eBook and book
- High royalties for the sales
- Completely free - with ISBN
- It only takes five minutes
- Every paper finds readers

Publish now - it's free