Sie sind auf Seite 1von 17

Research Training in Spain: An Assessment exerciseS.

Antonio García-Romero1, Aurelia Modrego2


1
Dept. of Economics. Universidad Carlos III de Madrid and UOC.
2
Dept. of Economics and Instituto Flores de Lemus. Universidad Carlos III de Madrid

Abstract

This paper analyses the effect of a postdoctoral fellowship on scientific productivity.


After reviewing the literature on research training, a model based on Human Capital
theory is proposed as an attempt to measure such an impact. Main results show that
scientific productivity is due both to the ability of Ph.D. graduates and to their
engagement on the postdoctoral position. However, we prove the scope of the former
impact becomes approximately seven times the associated to the later one. Therefore,
two recommendations for public policy design can be made. On one hand, it is
recommended: a) to improve the quality of Ph.D. programmes as the most effective
approach to enhance scientific productivity, b) to design a rigorous information
system about the quality of existing Ph.D. programmes and c) to define an adequate
grant program as a way to create a real European research training market. On the
other hand, it is necessary to enhance the selection process to postdoctoral fellowships
in order to achieve an efficient allocation of Ph.D. graduates to the appropriate
research groups.

Keywords: Human Capital; Research training; Postdoctoral Fellowship Programme; Structural


Equation Modelling.
JEL Classification: A23; J44; I21

S
This paper is part of Ph.D. Thesis of Antonio García Romero,with Aurelia Modrego as advisor. The
authors wish to thank to Juan J. Dolado, Carmelo Nuñez, Antonio Romero-Medina, for their comments
and support. Both authors also thanks to Dirección General de Investigación Científica y Técnica
(DGICYT) for funding this research [APC-93-097].
1. Introduction.

The aim of this paper is to analyse the effect of postdoctoral fellowship programmes
upon scientific productivity of researchers. We focus on a postdoctoral fellowship as the
research training period corresponding to the first year of professional career right after
Ph.D. graduation.

As stressed out by literature, there are a number of factors as postdoctoral fellowship


scholarship (Arora, 1993), age and life cycle (Diamond, 1984; Stephan, 1994), research
activity performance of department (Hogan, 1981), scientific co-operation (Stephan,
1996; Modrego, 1998), and quality of training or individual's abilities and skills
(Anderson, 1989; Buchmueller, 1999) with significant relation with scientific
productivity.

From a broad perspective, research training of scientists can be conceived as the process
that takes place during their entire professional life, as far as research activity can be
understood as a mean of training itself (Carlson, 1995; Martin, 1983 e Irvine, 1980)1.
However, research training is usually referred as the period of time in between Ph.D.
graduation and the completion of an own managed research project after a postdoctoral
fellowship. According to this, two phases can be differentiated: the pre-doctoral and the
postdoctoral one. During the former, abilities and relevant knowledge bases are
developed whereas, during the later, most significative gains are the incorporation of
scientists to so called elite or "invisible colleges" (Garvey, 1971) and the achievement
of less codified knowledge as it becomes crucial for an outstanding research
performance.

In order to assess the impact of each scientist's ability upon his scientific productivity,
two effects are to be considered. Firstly, a direct effect for which higher levels of
individual's ability are associated to higher levels of scientific productivity. Secondly,
an indirect effect for which more skilful scientists, with higher levels of previously
acquired training, can use that levels as a signal which allows them to get a postdoctoral
position at a higher recognised research programme. Consequently, since more skilful
scientists are incorporated in research centres and departments associated to higher
levels of research performance, it will eventually have an impact on their scientific
productivity2 (Garvey, 1971). Therefore, it becomes relevant to analyse to what extent
the research productivity of a scientist can be directly associated to its individual skills,
either natural or acquired during his Ph.D. research, or in contrast, it can be indirectly
associated to the postdoctoral position.

In the aim of differentiate both components of scientific productivity, we need to


scrutinize those factors affecting and to quantify those of relevance. In our previous
work, first issue has already been solved by adopting two models. First model was
based on Human Capital Theory (Becker, 1964) as it included both direct and indirect
effect of scientist ability upon productivity. Second model was based on Screening
Theory (Cohn, 1990) which refers only to the indirect effect. According to our previous
1
It becomes as a unique mechanism for the transmission of certain pieces of knowledge. Moreover, given
the fact that most of them are of a tacit nature (training in job).
2
According to this rationale, known as Cumulative Advantage, the quality of postdoctoral fellowship acts
as a signal in the labour market.

2
results, first model was proved to be best suited to explain our data. Therefore, we may
conclude that scientific productivity is due both to postdoctoral fellowship and to
scientist's ability, either natural or acquired during Ph.D. training

Our estimated parameters for the Human Capital theory based model reveal that
scientific productivity is mainly due to scientist's ability rather than to research training
received during postdoctoral fellowship. We prove the scope of the former impact
becomes approximately seven times the associated to the later one. This result
highlights the importance of Ph.D. programmes as they contribute to: 1) enlargement of
the knowledge base of general concepts of science and scientific culture; 2) building up
of specific knowledge base of scientific discipline; 3) improve the capacity of analysis
and appraisal of the codified knowledge, both at the scientific and the socio-economic
level; and 4) increase the capacity to propose and to carry out new research projects.
Along this line of reasoning, two recommendations for public policy design can be
made. On one hand, it becomes necessary to establish evaluation mechanisms for Ph.D.
programmes in order to get informed about their quality (Guimaräes, 1995) and,
subsequently, be able to provide financial support for scientists training at those
recognised institutions. Hence, a publicly accessible and rigorous information about the
quality of existing Ph.D. programmes and an adequate grant system will contribute to
create an efficient European research training market. On the other hand, it is required
that the selection process of applicants to postdoctoral fellowships will consider as a
criteria not only applicants ability but also other factors such as: 1) the eminence of
Ph.D. supervisor; 2) the research activity level of performance of Ph.D. programme
institution; and 3) the quality and relevance of research topic of Ph.D. project.

This paper draws on data collected from a sample of 200 Spanish Ph.D. graduates in
four scientific fields (Physics, Chemistry, Biology and Medicine) granted with a
postdoctoral position. Information was gathered on abilities of scientists and their
research activity outcomes. After identifying main research hypothesis, results have
been obtained by means of path analysis and confirmatory factorial analysis (Bentler,
1995; Kline, 1998) both techniques within the Structural Equation Modelling
methodological approach.

The rest of this paper is organised as follows. Section 2 includes a literature review on
research training. Section 3 presents theoretical models and equations to test. Section 4
describes data sources and methodological approach based on Structural Equation
Modelling. After presenting our main results in Section 5, conclusions and
recommendations are drawn in final section.

2. Literature review
In this section literature contributions are organised in four groups : a) qualitative
studies; b) descriptive studies ; 3) those focussed on the correlation between research
training and productivity ; and 4) those centered on the causality between research
training and productivity. In addition, a separated group is also included concerning
with recent research carried out in Spain.

3
Qualitative Studies. This group is mostly formed by studies commissioned by national
and supranational organisations (e.g. NSF, OECD). Based on aggregated data and
experts’ opinions, these studies focus on key problems regarding research training and
public policy options across different countries. In OECD (1995) a comparison among
research training systems of several countries is made in order to identify major
challenges and the way of dealing with them. Spain it was not included in this review. A
similar study has been carried out by the National Science Foundation (NSF, 1996 and
Smith III, 1995). The purpose was to identify weaknesses within the research training
system in USA. Main findings indicate that the excessive level of specialisation of
scientists makes them not very useful for the society out of the academic environment.
Other works have focused on countries with less significant research training tradition,
as it is in the case of Leta (1998) in Brazil and Stirati (1995) in Italy.

Descriptive Studies. This group includes studies based on descriptive analysis of


survey data gathered from scientists inquired about their research training period. Irvine,
(1980)3 collects information from 277 Ph.D. graduates in Astronomy. The consultation
pursued to inform on the relevance of individual's abilities, acquired by scientists during
the execution of the Ph.D. research, on their professional trajectory. Main findings
confirm that such abilities are of great utility both in the context of public and private
research activity. Halfpenny et al (1992) evaluates a research training grant scheme at
the UK, SERC Grants, a programme addressed both to Ph.D. and degree graduated
applicants (n = 468). Analysis is focus on the characterisation of professional activity
after research training and labour assignments, though no further research is made to
account for the relationship between both issues. A similar work was carried out for the
evaluation of the EC Training Fellowship Programme (Teichler, 1991a; 1991b). The
study collected opinion scientists involved in 472 approved applications and 140 cases
in which scientist rejected the scholarship. Main findings stressed out the imbalance in
terms of scientific level among EU countries: a) a greater rate of rejection is given when
receptor country's scientific level is lower; and b) both the impact upon research activity
and the overall degree of satisfaction is higher among those scientists with origin in
countries with lower scientific level.

Studies based on the correlation between research training and productivity. In


some of them review the relationship from a bivariante perspective as is the case of
Cole and Cole (1967), Crane (1965), Over (1982) and Zuckerman (1975). In Hogan
(1981); Long (1978); Evered (1987) and Anderson (1989), a multivariate approach is
followed although the causality relationship is not yet analysed to detail.

Studies based on the causality between research training and productivity. This
fourth group include literature concerned with any kind of causality link between
research training of scientists and their subsequent research activity. The underlying
concept is that the learning process takes place on a sequential mode and it is
determined by a number of factors such as individual's abilities, quality of training and
contextual framework (public policies or gender, among others). The casual approach is
better adjusted to real situations since it considers that individual's ability has an impact
on the output of research training, for which the scientist gain access to more recognised
departments and consequently, achieve higher degrees of learning. Besides, research
training has an impact on the quality level of the institution hosting the postdoctoral
3
This piece of research belongs to an evaluation of the scientific and economic effects of ’Big Science’.

4
position, as well as on the degree of its fulfilment. The quality of training during the
postdoctoral fellowship determine the final level of activity of the scientist. In addition
to this casual chain, a causality approach includes direct effects such as the relationship
between ability and scientific productivity. Most relevant studies along this reasoning
are Cole (1973), Chubin (1981), Helmreich (1982) and specially, Rodgers (1989) and
Maranto (1994). Main conclusions of these works are as follows. By using path
analysis, Cole and Cole (1973) obtain insights on the acquisition process of 'scientific
prestige' defined as reputation within department and clarity. By employing professional
prestige as a predicting variable, both as a combination of publications during early
professional career and publication citations as well as the quality of the Ph.D.
programme, they prove a significant effect of prestige of Ph.D. programme on scientific
productivity. However, they specify this effect is higher at the long term than for early
years of professional career after Ph.D. completion, either expressed in terms of
publications or citations.

In a similar way, several studies focus on the research activity of Psychologists. In the
first survey (Chubin, 1981) regression models are proposed to predict publications from
a) the level of quality of the department of the Ph.D. programme, b) a postdoctoral
fellowship scholarship c) the Ph.D. project supervisor, d) the number of publications
before Ph.D. thesis completion and number of publications resulting from Ph.D.
research. Main conclusion is that the unique statistically significant predictor of
publications is the number of publication citations achieved before pre-doctoral phase.
The study performed by Helmreich et al. (1980) considers gender and scientists'
personal motivations as predicting variables. Results show that personal motivations
have positive impact upon results; male scientists carry out their Ph.D. projects in
highly scored departments, work in institutions of higher prestige, publish a higher
number of papers and receive a higher level of citations4.

An interesting contribution is the study presented by Lee Hansen and Weisbrod (1978)
in which it is not only considered the effect upon scientific productivity but also the
earnings of scientists. McGinnis, Allison y Scott Long (1982) analyse in detail the effect
of postdoctoral fellowship upon scientific productivity. They conclude that such an
effect is relatively small. Scott Long y McGinnis (1985) perform an exhaustive analysis
of the effect that a mentor has on the professional career of Ph.D. graduates. In an
interesting contribution, Buchmueller, Dominitz y Lee Hansen concentrate on the early
productivity of PhD Economists (1999) bringing out new variables into play such as a
research assistantship during Ph.D. research. Just recently, Magematin (2000) has
addressed this topic from the perspective of the Ph.D. labour market by making use of
the Human Capital theory.

Finally, the work carried out by Rodgers (1989) and Maranto (1994) are relevant both
because their methodological approach, based on Structural Modelling Equation, and
because their theoretical base as they take into account individual's ability and its
potential effect on scientific outcomes from various aspects. From a multidisciplinary
perspective, the work by Rodgers (1989) attempts an analysis on productivity from a
sample of Physiologists. Main novelty comes from the incorporation of individual's
ability as an explaining variable which permits to avoid the bias created by the
individual differences (Cohn, 1994), a common issue in this type of approaches. A
4
Known as ’Matilda effect’ (Rossiter, 1993).

5
comparison between Human Capital theory and Cumulative Advantage theory is
presented by Maranto (1994) in order to explain scientific production in the discipline
of Accounting. He suggests a sophisticated model by using a measuring model known
as hybrid model (Kline, 1988) as it simultaneously considers two type of variables:
latent and observable ones.

Studies based on the Spanish case. To end up this review, a selection of works on
research training for the Spanish case are provided. In Martín (1996) and Martín-
Sempere (1998), data of scholarships abroad from the period 1984-1994 is presented
and classified by scientific fields and country of destination. López Aguado (1987)
compares scientific production of Ph.D. graduates in Physics during the academic year
1976-77 with the total scientific production in that field in Spain. An outstanding level
of research production in international terms is revealed both within the recent Ph.D.
graduates and the total sample of scientists. A final work selected is the performed by
Fernández Esquinas (1995) on a base of a sample of Ph.D. students and Ph.D. graduates
from all scientific fields (n = 1,200). Respondents were asked to provide detailed
information regarding their scholarship (amount of money received, means and time
among others). Data exploitation indicate that professional trajectories are highly
concentrated in the academic system as a result of high levels of embedding. Nearly
53% Ph.D. graduates follow their professional career in their Ph.D. programme
institution. He also provides us with useful data regarding research activity of scientist
during and after the scholarship.

3. Hypotheses

This section describes, in brief, main hypothesis of the Human Capital Model in order
to establish the relationship between ability and scientific productivity.

Many applications in different contexts have been used since Becker (1964) presented
the postulates of this theory. In this sense, some researchers have applied it with relative
success in order to explain how scientific productivity evolves during scientist's life
cycle (Diamond, 1986; Weiss, 1982; Levin, 1991 and Siow, 1994). According to this
theory, training period is seen as an investment during which an individual decides to
continue being trained as far as the cost associated to it remains lower than the profits he
or she expects to obtain after that period. Thus, an individual would resolve to be
trained until the marginal cost associated with an additional period of training is equal
to the marginal benefit that he or she expects to obtain afterwards. Since training period
permits to acquire relevant knowledge base this will have a direct effect on productivity.
Additionally, since the most skilful individuals are the most productive ones and also
those capable to learn more, a double relationship of individual's ability and
productivity can be appreciated. Furthermore, there will be an indirect effect through
training.
We attempt to apply this theory to research training that takes place during a
postdoctoral fellowship programme taken up abroad. Our approach is justified by the
fact that it is a way of in-job training and, therefore, a type of Human Capital.
Hypothesis underlying the Human Capital approach are as follows:

6
Hypothesis 1.1: There is a positive relationship between individual's abilities and the
quality of research training received.
Hypothesis 1.2: There is a positive relationship between individual's abilities and
productivity.
Hypothesis 1.3: There is a positive relationship between the quality of training and
productivity.

According to it, individual's ability (HAB) has both an impact on the quality of the
research training they receive (EST) as well as a direct effect on productivity (PRO).
Besides, productivity is also influenced by the quality of research training (EST).

4. Data and methodology

4.1. Data
This analysis draws on the extensive empirical work undertaken in 1994 for the
evaluation of the MEC-Fleming Postdoctoral Fellowship Programme5 (García Romero,
1996). The population was formed by all Spanish applicants to a postdoctoral position
taken up abroad during the period 1985-1992 (N = 1,679). A sample of 419 scientists
from all fields was postal surveyed. The response rate was quite good as 267 scientists
satisfactory returned the questionnaire, out of whom 228 sent an attachment with their
Curriculum Vitae which ended up to be a valuable information. Details are displayed in
table 1. Responses from Humanities and Social Sciences scientists were dropped out
since data revealed a different behaviour as compared with the observed for
Experimental Sciences individuals and represent a small proportion (28 cases). Final
sample consisted of a group of 200 cases involving an error in estimations of about 4%.

Table 1. Population and sample


Group No. (%)
Population 1,679
Sample 419 (100)
Total survey responses 267 (63.72)
With Curriculum Vitae 228 (54.41)
With Curriculum Vitae with cert. 200 (47.73)

With regard with similar works on this topic, the questionnaire we developed for this
survey presented important advantages as follows:

1. Not only successful applicants to the postdoctoral fellowship programme are


considered but also rejected applicants. This approach will permit to compare

5
Two grant schemes were open to applicants during this period of time. Fist mechanism was the MEC-
Fleming programme accounting for 56% of applications and 58% of approvals. Co-organised by the
Ministry of Education and Science and the British Council, it consist of a specific postdoctoral
fellowships at UK institutions. The second mechanism was named as General Programme, it supported
postdoctoral positions to be taken up in foreign institutions, including UK institutions. However their
similarities, MEC-Fleming was associated to higher quality of research and applicants hold higher
profiles. This fact allows us to define a variable which differentiates each type of postdoctoral fellowships
in terms of their level of quality.

7
both cases and identify the additionally associated to the postdoctoral fellowship
programme as well as those factors of relevance in the selection process.
2. Questions on professional trajectory after completion of postdoctoral position
were included as well as annual incomes. This information will allow us to
perform further research in the aim of explaining postdoctoral fellowship impact
on professional career of scientist.
3. The request of an attached Curriculum Vitae contributed with reliable data about
individuals scientific production and professional career of great interest for
exploitation.

4.2. Methodology
We have estimated each model using Structural Equation Modelling, SEM, which
implies a model of simultaneous equations with observable variables, known as path
analysis, and measurement models as confirmatory factorial analysis (CFA). The
combination of both techniques is deemed suitable for our research objectives as it
allows us to obtain hybrid models which permit to estimate parameters, test the godness
of fit and construct latent variables when more than one indicator of such variables are
available. For this task we have employed the statistical software EQS (Bentler 1995).

4.3. Variables and impact measures

Those variables included in the proposed hybrid models based on different theories are
not directly observable. This is the case for individual's ability and scientific
productivity, which are latent variables. This implies the need to work with indicators6.
In following sections, the indicators related to productivity, postdoctoral fellowship and
individual's ability are described.

Scientific productivity measures (PRO).

Research activity conveys different aspects that lead to scientific outcomes of different
nature. For our purpose, two kinds of measures related to scientific publications are
considered as they refer to: 1) those publications in journals included at the Science
Citation Index (SCI); and 2) those publications in journals not included at the SCI,
which at first, are considered as less demanding journals7. On the basis of data obtained
from the CVs, the following measures were settled:
Scientific productivity 1: Number of publications per year in journals included at the
SCI (PISID).
Scientific productivity 2: Number of publications per year in journals non included at
the SCI (PNISID).
Scientific productivity 3: Average impact factor of SCI journals (FID)

From all of these measures, the number of publications in journals included in the SCI
is the one most widely used among similar works given the availability to develop
6
In Martin and Irvine (1983) the parcial character of indicators regarding scientific activity and outcomes
is stressed out. The use of several indicators for each variable is recommended. Besides, Martin (1996)
concludes in his survey that, in the majority of cases, one or two indicators at most are used for each
variable.
7
The ISI includes journals on the base of their average impact per each received citation.

8
reliable measures such as citations and impact factor. For our purpose, estimations are
made separately considering whether the scientific output is published in a journal
included in the SCI or in a journal not included. We only include the impact factor for
the first type of journals.

Research training during postdoctoral fellowship measures (EST).

Bearing in mind that our sample consist of those scientists granted with a postdoctoral
fellowship and those whose application was withdrawn at the selection process8, a
variable with four modalities was created in order to characterise the quality of the
postdoctoral fellowship: 1) no postdoctoral fellowship; 2) short period postdoctoral
fellowship; 3) one year postdoctoral fellowship under MEC-General Programme; and 4)
one year postdoctoral fellowship under MEC-Fleming Programme.

Individual's ability measures (HAB).

Individual's ability is characterised by a measuring model built up from two indicators:


1) number of publications in SCI journals during the training period of the Ph.D.
programme; and 2) average impact factor associated to those publications. Two reasons
account for this selection: firstly, both are measures closely related to research activity
and consequently, to the necessary individual's ability to become a recognised scientist;
and secondly, the analysis of the correlation matrix reflects convergent validity and
therefore, both can be considered measures of the same latent variable, see table 3. For
this reasons, scientific production during Ph.D. programme (before postdoctoral
fellowship) can be considered as a proxy of individual's ability. Consequently, the
indicators are:

Ability 1: Number of publications per year in SCI journals before postdoctoral


fellowship (PISIA).
Ability 2: Average Impact Factor of journals before postdoctoral fellowship (FIA)

8
Without discriminating by factors such as country of destination or organisational context of the position
(e.g. whether it is in HEI, industry, Spain or abroad), evaluation criteria includes factors such as those
regarding individual characteristics of scientists; recognition of host institution of postdoctoral position;
and relevance of research to perform.

9
5. Results

5.1. Descriptive statistics


Table 2 presents the means and standard desviations of all variables. As expected, data
reflects that an increment on publications in SCI journals as well as on their average
impact factor takes place after postdoctoral fellowship.

Table 2. Descriptive statistics


Variable Mean S.D. Definition
PISIA 0.714 0.736 No. SCI Publication/year before postdoctoral fellowship
FIA 1.324 1.286 Average IF before postdoctoral fellowship
EST 2.460 1.181 Quality of postdoctoral fellowship
PISID 1.628 1.607 No. SCI Publ./year after postdoctoral fellowship
FID 1.841 1.854 Average IF after postdoctoral fellowship
PNISID 0.644 1.003 No. Non SCI Publ./year after postdoctoral fellowship

Following table presents the correlation matrix. It shows how variable PNISID presents
a negative correlation with the rest of variables with exception of PISID. At first glance,
publishing in non included SCI journals appears to be not such a common practice
among recognised scientist in Experimental Sciences.

Table 3. Correlation matrix


PISIA FIA EST PISID FID PNISID
1.000
0.220 1.000
0.079 0.195 1.000
0.410 0.101 0.173 1.000
0.159 0.535 0.185 0.172 1.000
0.004 -0.156 -0.039 0.149 -0.181 1.000

5.2. Modelling interpretation.


This section presents the estimations obtained from the Human Capital theory based
model as it is the model that best fit our data.

Human Capital theory based model (SCI)

Table 4 presents the estimations for the standarised coeficients within the Human
Capital theory based model obtained from SCI data. Results are consistent with the
theoretical hypothesis described in section 3.1.

10
Table 4 Human Capital Model (SCI)
Parametre (β) Hypothesis Value (t-stat)
HAB → EST + 0.283 ** (2.780)
HAB →PISID + 0.225 ** (2.230)
HAB → FID + 0.558 ** (3.610)
EST → PISID + 0.109 * (1.467)
EST → FID + 0.020 (0.265)
PISID → FID + 0.035 (0.482)
( ** ) coeficient estimates that are significant at the 5% level
(*)coeficient estimates that are significant at the 10% level

This type of equations allow us to split total effect (ET) of individual's ability (HAB) on
SCI productivity (PISID), as the sum of direct effect (ED =HAB → PISID) and indirect
effect (EI =HAB → EST × EST → PISID). Therefore, we may conclude:

ET = ED + EI = 0.225 + (0.283 × 0.109) = 0.225 + 0.031 = 0.256


Being the ratio between both effects as
ED HAB → PISID
= ≈ 7.25
EI HAB → EST × EST → PISID

we may conclude that the main factor affecting scientific productivity is Ph.D.
graduates' ability, being calculated such a contribution about 70% since
PISID = 0.225 × HAB + 0.109 × EST + ε
Previous results will allow us now to perform a sensibility analysis in order to compare
between those effects associated to: a) a increase in the level of quality of postdoctoral
fellowship while mantaining constant Ph.D. graduates' ability and b) an enhancement of
Ph.D. graduates' ability with constant level of postdoctoral fellowship. The following
table shows the effects on productivity associated to each situation.

11
Table 5 Effect on productivity SCI
Policy Increase on SCI productivity (S.D.)
(1)
EST + 1 std 0.109
HAB + 1 std(2) 0.225 + 0.031 = 0.256
EST + 2,35 std(1) 0.256
(1) Increasein the level of quality of postdoctoral fellowship with constant
Ph.D. graduates' ability
(2) Enhancement of Ph.D. graduates' ability with constant level of quality of
postdoctoral fellowship

Data reveals an equal effect between 1 S.D. enhancement of individual's ability of


scientists and 2.35 S.D. leverage of postdoctoral fellowship. This stresses out the
relevance of the first choice as a public policy opcion against the second one.

Human Capital theory based model (non SCI)

Table 6 contains standarised parameters for the Human Capital theory based model of
scientific outcomes not included at the SCI. As it happens in the previous case,
coeficients confirm theoretical hypothesis described in section 3.1 but with the
exception of the relationship between individual's abilities and publications in journals
not included in the SCI.

Table 6. Human Capital theory based model (non SCI)


Parameter (β) Hypothesis Value
HAB → EST + 0.241**(3.487)
HAB → PNISID – -0.174**(-2.414)
EST → PNISID + 0.04 (0.052)
(**) coeficient estimates that are significant at the 5% level

Given the parameter for the direct impact of postdoctoral fellowship on scientific
productivity (EST → PNISID = 0), we obtain ET = ED + EI = −0.174 + 0 , therefore,
higher levels of individual's abilities correspond with lower level of scientific outcomes
among publications in journals not included in the SCI. Not surprisingly, this result is
consistent with the intuition for which most skilfull scientists are those who publish less
in journals not included in the SCI.

12
6. Conclusions and recommendations.

The influence that research training received by scientists has on their subsequent
scientific productivity is a crucial issue for the definition of Science and Technology
Policy. This present study has been based on previous developments by Rodgers (1989)
and Maranto (1994) in which Structural Equation Modelling approach and their
proposed measures are proved to be suitable to analyse the impact of postdoctoral
fellowships on scientific productivity.

Main findings can be summarise as follows. Firstly, results confirm that Human Capital
theory is best suited to explain scientific productivity than Screening theory in the
context of this analysis. It implies that individual's ability has a significative impact on
scientific output. In addition, it is worth noting that models present better godness of fit
indexes when attempting to explain scientific production via publications in journals not
included at the SCI. Secondly, the estimations of parameters reveal that scientific
productivity is mostly due to individual's ability developed during the Ph.D. research
training rather than during the work at the postdoctoral position. We prove the scope of
the former impact becomes approximately seven times the associated to the later one.

This results underline the necessity to consider the training of scientists as a learning
process with two differentiated phases. In the first one, corresponding to the Ph.D.
programme training, it has been proved that abilities acquired in it have a significant
effect on the scientific productivity of researchers. Other characteristics of great benefit,
both in the context of public research and in the private one associated to this phase are:
the enlargement of the knowledge base of general concepts of science and scientific
culture and the acquired capacity of analysis and appraisal of the codified knowledge,
both at the scientific and the socio-economic level.

Alongside with this line of reasoning, two recommendations for public policy design
can be made. On one hand, it becomes necessary to establish evaluation mechanisms for
Ph.D. programmes in order to get informed about their quality and, subsequently, be
able to provide financial support for scientists training at those recognised institutions.
A publicly accessible and rigorous information about the quality of existing Ph.D.
programmes and an adequate grant system will contribute to create an efficient
European research training market. The implementation of an assessment system for
Ph.D. programmes, as it is proposed by the NRC, would permit students to have full
information access to each programme offer and subsequently, it would set up the frame
conditions for an authentic European Research Area.

On the other hand, data reveals that postdoctoral fellowships is also positively related
with scientific activity. In addition, the indicator used in this work suggests that it could
be convenient that the selection process of applicants to postdoctoral fellowships will
consider not only applicants' ability but also other factors such as the eminence of Ph.D.
supervisor, the level of research activity performance of Ph.D. programme institution
and the quality and relevance of research topic of Ph.D. project. In this sense, the
allocation of highly skilled Ph.D. graduates to a lower level of quality group or the
contrary situation, will have long term consequences on scientific productivity.

13
Finally, on the base of our results, it could be very useful for the policy-making process
to ensure the availability of data regarding scientists' ability that will allow to perform
analysis of this kind in order to obtain interesting conclusions from a policy-making
point of view.

7. References

Anderson, J. 1989. The evaluation of research training. In Evered, D (ed). The


Evaluation of Scientific Research. 93-113. Chichester: John Wiley & Sons.
Arora, A, and A Gambardella. 1996. “The Impact of NSF Support for Basic Research
in Economics“ Working Paper (ewp-othr/9702001). Carnegie Mellon University.
Bayer, AE, and JE Dutton. 1977. Career Age and Research Professional Activities of
Academic Scientists. Journal of Higher Education XLVIII, no. 3 (MAY/JUNE):
259-82.
Bentler, P. 1995. EQS Structural Equations Program Manual. Encino, CA: Multivariate
Software, Inc.
Becker, GS. 1964. Human Capital. New York. National Bureau of Economic Research.
Bollen, KA. 1989. Structural equations with latent variables. New York:John Wiley &
Sons.
Buchmueller, TC, J Dominitz and W Lee Hansen. 1999. Graduate Training and the
early career productivity of PhD. economists. Economics of Education Review; 14,
65-77.
Carlson, T, and D Martin-Rovet. 1995. The Implications of Scientific Mobility
Between France and United States. Minerva 33: 211-50.
Chubin, DE, AL Porter and ME Boeckmann. 1981. Career Patterns of Scientists: A
case for Complementary Data. American Sociological Review, 46, 488-498.
Cohn, E and T.G. Geske. 1990. The Economics of Education. New York: Pergamon
Press.
Cole, J and S Cole. 1973. Social Stratification in Science. Chicago: University of
Chicago Press.
Cole, S and J Cole. 1967. Scientific output and recognition: A Study in the operation in
the reward system in science. American Sociological Review, 32, 377-390.
Crane, D. 1965. Scientists at major and minor universities: A Study of productivity and
recognition. American Sociological Review, 30, 699-714.
Dasgupta, P, and PA David. 1994. ”Towards a New Economics of Science.” Research
Policy : 487-521.
Diamond Jr, AM. 1984. An Economic Model of the Life-Cycle Research Productivity
of Scientists. Scientometrics 6, no. 3: 189-96.
Evered, DC, J Anderson, P Griggs, and R Wakeford. 1987. The Correlates of
Research Success. British Medical Journal 295: 241-6.

14
Fernández Esquinas, M, MT González de la Fe y M Pérez Yruela. 1995. La
Formación de científicos en el Plan Nacional de I+D (1982-1994): Una
aproximación evaluativa. Documento de Trabajo 95-14. Córdoba. IESA.
García-Romero, A. 1996. Evaluación del Subprograma de Becas Postdoctorales MEC-
Fleming. Informe final. Instituto de Estudios Avanzados en Economía ’Flores de
Lemus’.
Gardfield, E. 1996. Fortnightly Review: How Can Impact Factors Be Improved?
British Medical Journal 313: 411-3.
Garvey, WD, and BC Griffith. 1971. Scientific Communication: Its Role in the
Conduct of Research and Creation of Knowledge. American Psychologist 26: 349-
62.
Gelso, CJ, B Mallinckrodt and A Brust Judge. 1996. Research Training
Environment, Attitudes Toward Research and Research Self-Efficacy: the Revised
Research Training Environment Scale. The Counselling Psychologist 24, no. 2:
304-22.
Guimaräes, J. and Humann, M. 1995. Training of human resources in Science and
Technology in Brazil: the importance of a vigorous post-graduate program and its
impact on the development of the country. Scientometrics 34, 101-119
Halfpeny, P, C Pettipher and L Georghiou. 1992. A Study of Research Assistants
Supported by SERC Grants. PREST. University of Manchester.
Helmreich, RL and JT Spence. 1982. Gender differences on productivity and impact.
American Psychologist, 36, 1142.
Helmreich, RL, JT Spence, WE Beane, GWLucker and KA Mathews. 1980.
Making it in academic psychology: demographic and personality correlates of
attainment. Journal of Personality and Social Psychology, 39, 896-908.
Hogan, TD. 1981. Faculty Research Activity and the Quality of Graduate Training. The
Journal of Human Resources XVI: 398-415
—1985. The Publishing Performance of U.S. PhD. Programs in Economics During the
1970s. The Journal of Human Resources no. : 216-29.
Irvine, J, and BR Martin. 1980. The Economic Effects of Big Science: the Case of
Radio Astronomy. Proceeding of the International Colloquium on the Economic
Effects of Space and Other Advanced Technologies. Paris: ESA.
Kline, RE. 1998. Principles and Practice of Structural Equation Modelling. New York:
Guilford Press.
Kostoff, RN. 1997. Citation Analysis Cross-Field Normalization: a New Paradigm.
Scientometrics 39, no. 3: 225-30.
Lee Hansen, W, and BA Weisbrod. 1978. "Modelling the Earnings and Research
Productivity of Academic Economists." Journal of Political Economy 86, no.
41:729-41.
Leta, J, D Lannes, and L De Meis 1988. Human Resources and Scientific Productivity
in Brazil. Scientometrics 41, no. 3: 313-24.

15
Levin, SG and PE Stephan. 1991. Research Productivity Over the Life Cycle.
American Economic Review, 81(1), 114-32.
Long, JS. 1978. Productivity and Academic Position in the scientific career. American
Sociological Review, 43, 889-908.
López Aguado, G. 1987. El doctorado Español en Física a Través de sus Publicaciones.
Revista Española de Física. 1(1). 28-35.
McGinnis, R, PA Allison, and JS Long. 1982 "Postdoctoral Training in Bioscience:
Allocation and Outcomes." Social Forces 60, no. 3: 701-22.
Mangematin, V. 2000. PhD Job Market: Professional Trajectories and Incentives
During the PhD. Research Policy 29, no. 6: 741-56
Maranto, CL, and CA Streuly. 1994. The Determinants of Accounting Professors’
Publishing Productivity - The Early Career. Contemporary Accounting Research
10, no. 2: 387-407.
Martin, BR. 1996. The Use of Multiple Indicators in the Assessment of Basic
Research. Scientometrics 36, 3: 343-62.
Martín, MJ, LM Plaza, and J Rey. 1996. Transnational Mobility of Researchers: the
Spanish Fellowships Programme for ’Senior’ Scientists. Research Evaluation 6, no.
1: 13-18.
Martín-Sempere, MJ, LM Plaza, and J Rey. 1998. La Movilidad Temporal de los
investigadores doctores en el Extranjero. Política Científica: 50-53
Martin, BR, and J Irvine. 1983. Assessing Basic Research. Some Partial Indicators of
Scientific Progress in Radio Astronomy. Research Policy 12: 61-90.
Modrego, A. 1998. El español y la Ciencia. Jornadas sobre el español. Instituto
Cervantes y Fundación Duques de Soria. Tordesillas.
NSF (National Science Foundation). 1996. Higher Education in Science and
Engineering. US Science and Engineering in a Changing World. NSF, 2-1/2-35.
Washington, DC: NSF-SRS.
OECD. 1995. The Future of Research Training. Paris: OECD.
Over, R. 1982. Research productivity and impact of male and female psycologist.
American Psycologist, 37, 24-31.
Rodgers, RC, and CL Maranto. 1989. Causal Models of Publishing Productivity in
Psychology. Journal of Applied Psychology 74, no. 4: 636-49.
Rossiter, MW. 1993. The Matilda Effect in Science. Social Studies of Science, 23, 325-
341.
Seglen, PO. 1997. Why the Impact Factor of Journals Should Not Be Used for
Evaluating Research. British Medical Journal 314: 497.
Schwartz, S, and J López Hellin. 1996. Measuring the Impact of Scientific
Publications. The Case of Biomedical Science. Scientometrics 35, no. 1: 119-32.
Shubert, A, and T Braun. 1996. Cross-Field Normalization of Scientometric
Indicators. Scientometrics 36, no. 3: 311-24.

16
Siow, A. 1994. The Organization of the Market for Professors. Workshop in
Applications of Economics.
Smith III, TP, and JC Tsang. 1995. Graduate Education and Research for Economic
Growth. Science 270, no. 6 October: 48-9.
Stephan, PE, and SG Levin. Measures of Scientific Output and the Age-Productivity
Relationship. In: Handbook of Quantitative Studies of Science and Technology.
van Raan (ed.). North Holland. Amsterdam.1988.
Teichler, U. 1991a. Evaluation of the EC Training Fellowship Programme Based on a
Fellow's Questionnaire Survey. Scientometrics 21, no. 3: 343-65.
Teichler, U, G Lewinson, and L Massimo. 1991b. Surveys of European Community
Transnational Research Fellows. Research Evaluation 1, no. 3: 137-47.
Weiss, Y and L Lillard. 1982. Output variability, academic labour contracts, and
waiting times for promotion. In Research in Labour Economics.RG Ehrenburg (ed).
157-188.
Zuckerman, H and J Cole. 1975. Women in American Science. Minerva, 13, 82–102.

17

Das könnte Ihnen auch gefallen