1

The Determinants of Research Productivity: A Case Study of Mexico

Claudia Gonzalez-BrambilaCarnegie Mellon University

Advisors: Francisco Veloso and Granger Morgan

2

Agenda

Introduction Objective The Data The Model Results Conclusions

3

Introduction There is a general consensus that S&T

investments enhance economic growth, health, education and security.

However, there is an increased necessity to assess the impact of S&T programs.

This situation has shaped a culture of evaluation and monitoring in research. Publications and citations are the most

common ways to measure the importance of the contributions of a researcher or an organization.

4

Introduction (cont)

Publications play an important role: To get access to financial resources for research To recruit and give raises and promotions to scientists To rank universities’ departments As a channel of communication with industry

However, there are not many studies that explore the determinants of individual and collective research productivity

Implications for university administrators and policy makers

Forecast the expected productivity of faculty Design policies to enhance productivity Parameters to compare researchers across different

areas of knowledge and across countries

5

Main Findings of Studies Related to the Determinants of Research Productivity

Age Life cycle with peak between 40 and 50

Gender Women publish between 26 to 91% less than men

Education Graduates from top schools, with RA experience and employed in research universities are more productive. Graduates from Grande Ecoles are more productive

Cohort effect

No evidence

Countries studied United States, France and Israel.

6

Objective To explore the determinants of research

productivity using a data base of the most productive researchers in Mexico. Age Gender Areas of Knowledge Cohort Country of PhD Level in SNI Critical Mass Budget

7

Importance

This is the most comprehensive study of its kind. Number of researchers Considers all areas of knowledge The first outside the developed world

8

Data Base Information on 14,328 researchers

in the SNI from 1991 to 2002. The data are classified in two

categories:1. Characteristics of the researchers:

Age Gender Country where PhD was earned Area and discipline

9

Data Base (cont.)

2. SNI variables: Level in the system for each year (or no presence) Budget of Conacyt Total number of researchers in SNI by area and

discipline Total number of publications by SNI researchers by

area and discipline.

The source of publications is the Science Citation Index produced by the ISI.

Publications per year per researcher Citations per 4 years per publication

10

Number of publications per SNI researcher by year and area

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1975 1980 1985 1990 1995 2000 2005

Year

Me

an

of p

ub

/ye

ar

All areas

Exact Sc

Biol. & Chem.

Health Sc

Social and Hum.

Biotech & Agric.

Engineering

11

The model Publicationsit = F (Xit, Zi, ci, uit)

Zi: Stable across time but not across researchers area, gender, country of PhD, cohort

Xit: varies in both dimensions age, age2, level in SNI, lagged budget, total number of researchers, total

number of publications. ci: individual unobserved effect uit: unobserved effect

Method First step: Negative binomial fixed effect model

To estimate the parameters of Xit Second step: Non linear least square method

To estimate the parameters of Zi

Sample Researchers with at least one publication between 1991-2001.

N=7,793 To adjust for quality, citations per 4 years. Publications between 1991-

1997. N=5,658

12

Results - AgeResearch Productivity Over the Life Cycle: All Areas of Knowledge

Age

02

Pu

blicatio

ns p

er Year

20 40 60 80 100

13

Results - Age (cont.)

Area of Knowledge

Researchers are productive between: (years old)

Peak of publications per year

The peak of productivity is at age: (years old)

All 29-77 1.5 57

Exact Sc. 34-82 1.6 62

Biol. & Chem. 25-83 2.4 58

Health Sc. 28-78 2 56

Social & Human. 35-72 0.9 58

Agric. & Biotech 33-72 0.8 56

Engineering 32-70 1.2 55

Disciplines

Physics 34-82 1.6 62

Biology 25-82 2.7 58

Materials Eng 29-75 1.5 57

14

Comparison with Other Studies (Physics)

Factor Mexico USA (Levin & Stephan)

France (Tuner & Mairesse)

Peak of publications / year

1.6 2 2.9

Decline starts (age)

62 45 52

Women publish (paper on average / year)

-0.16 - -0.9

15

Possible Explanations SNI encourages researches to continue publishing

The base salary in Mexico is about one third of what researchers actually receive.

Cannot easily be tested because of the lack of an adequate control sample.

The eldest researchers of a research group tend to appear as coauthors in the publications of their colleagues.

Will be tested in future research. Mexican researchers start and finish their careers

some years later than their colleagues in other countries.

There are no systematic differences between those who got their PhD when they were 30 or younger and those who were above 30.

16

Results - Gender Women publish 0.08 papers less

than men on average per year

Area of knowledge Proportion of female (%) Gender difference in scientific productivity (paper per year)

Exact Sc. 15 -0.16

Biology & Chemistry 36 -0.08

Health Sc. 39 -0.24

Social & Humanities 38 0

Agricultural Sc. & Biotech. 22 0

Engineering 12 0.06

17

Other Results Cohort effects

Researchers who got their PhD before 1960 are more productive than the latest educated.

The latest educated are more productive than researchers in cohort 1971-1980 and 1981-1990

Country of PhD Researchers who got their PhD in the US or Europe are slightly less

productive but graduate when they are younger. Level in SNI

The big jump in research productivity is seen when researchers change from candidate to level 1.

Critical Mass Small but positive effect in the total number of publications in the

same area of knowledge Little effect of saturation considering the total number of researchers

in the same area of knowledge Budget

Positive relation between the budget of the previous year and publications

18

Results for citations The peak of citations per four years is 0.3 and

is reached when researchers are 58 years old; this is only one year older than what was found for publications.

Publications of researchers in level 3 receive more citations than any other level.

the most important difference among levels in the program is seen in researchers in Exact Sc.

Health Sc. receive the largest number of citations per four years.

No important difference in what was found with publications

19

Conclusions The effect of age in research productivity is

not very important. The SNI might have enhanced research productivity

until the end of researchers working lives. Significant differences among areas of

knowledge. The design of public policies could respond to these

differences. There is no important gender gap in research

productivity. Potentially cohort effects.

20

Questions/comments

21

Distribution by Area of Knowledge

Engineering16%

Exact Sc.16%

Biology and Chemistry

17%

Health Sc.10%

Social and Humanities

25%

Agricultural Sc. and Biotechnology

16%

22

Mexican Publication in ISI

0

1000

2000

3000

4000

5000

6000

7000

Nu

mb

er

of

Pu

bli

ca

tio

ns

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

Year