Professional Documents
Culture Documents
Castilla, E. J. (2009) - The Institutional Production of National Science in The 20th Century. International Sociology, 24 (6), 833-869.
Castilla, E. J. (2009) - The Institutional Production of National Science in The 20th Century. International Sociology, 24 (6), 833-869.
Castilla, E. J. (2009) - The Institutional Production of National Science in The 20th Century. International Sociology, 24 (6), 833-869.
abstract: Science and scientific production have been widely promoted as power-
ful tools for advancing national economic and social development. While much
progress has been made in determining whether this is the case, less understood
are the underlying factors influencing national scientific activity in the first place,
especially during its 20th-century global expansion. In order to advance our
understanding of the development of science and world polity, this study investi-
gates in-depth when and under what functional and institutional conditions
countries chose to join any of the scientific unions comprising the ICSU, the pre-
eminent and oldest international science institution in the world. According to
analyses of historical data for 166 countries from 1919 to 1990, functional argu-
ments are only important predictors of the rate at which nation-states join scien-
tific organizations early in the ‘science diffusion’ process. After 1945, institutional
factors best account for worldwide national scientific activity: The joining rate
increases more quickly during the post-Second World War era with the rise of the
world system. This article also provides evidence of both convergence in the evo-
lution of national scientific activities and of the great invariability in the impact of
functional and institutional factors for core and peripheral countries over time.
The article concludes by discussing the implications of this research for the future
study of national scientific production and development in the world.
833
International Sociology Vol. 24 No. 6
range of social sectors and areas such as the economy (Meyer et al., 1975),
education (Drori and Moon, 2006; Frank and Meyer, 2007; McEneaney
and Meyer, 2000; Meyer, 1977b; Meyer et al., 1992; Ramirez, 2006; Ramirez
and Boli, 1987; Schofer and Meyer, 2005), employment law and practices
(Dobbin and Sutton, 1998; Dobbin et al., 1993), health care (Castilla, 2004;
Scott et al., 2000), law (Edelman, 1992; Edelman et al., 1999), human rights
(Boli, 1987; Hafner-Burton and Tsutsui, 2005; Tsutsui, 1999; Tsutsui and
Wotipka, 2004), women’s rights (Berkovitch, 1993; Charles, 1992; Ramirez
and Weiss, 1979), science (Wuthnow, 1980, 1987), the environment (Frank,
1992, 1997; Meyer et al., 1997; Schofer and Hironaka, 2005), welfare mod-
els (Strang and Chang, 1993) and models of management (Guillén, 1994;
Sahlin-Andersson and Engwall, 2002).
However, it has recently been argued that such convergence in national
institutions has tended to be both underestimated and undertheorized in
the social sciences (Drori et al., 2003: ix). Societies supposedly converge
as they selfishly strive for efficiency (as well as economic and political
dominance) and create institutions adapted to industrial modern econo-
mies (Inkeles and Rossi, 1956; Kerr et al., 1960; Rostow, 1960; Tönnies,
1963). However, in reaction to these functionalist lines of thought analyz-
ing nation-states as isolated actors evolving independently of other
nation-states, neoinstitutionalist theorists have continued the conver-
gence debate, arguing for convergence among industrial nations in areas
such as economic life, political systems, education, technology and strati-
fication systems (e.g. Baum, 1974; Drori, 2005; Grew, 1984; Jang, 2000;
Meyer, 1980; Meyer et al., 1975).
Science, scientific production and scientific experts are seen as central
to nation-states in two significant ways. First, their effects on national
economic development have been well supported by empirical research
(see, for example, Schofer et al., 2000). Even if that was not the case, as
some science professionals, nation-states and governmental organizations
believe, several cognitive models and frameworks have connected scien-
tific activity to societal development and progress (Schofer, 1999). Second,
science and scientific production are among the most important means by
which world culture is transmitted to and installed within particular
national societies (Drori et al., 2003). Not surprisingly, the rise of science
associations in the international sphere has had important theoretical and
practical implications for the overall expansion of scientific authority, as
well as for the scientization of many domains of society since the 1870s
(Schofer, 1999).
The purpose of this article is to contribute to this important research
stream by examining the functional and institutional factors influencing
global national scientific activity throughout the 20th century. In order to
do so, I empirically investigate when and under what conditions countries
834
Castilla The Institutional Production of National Science
835
International Sociology Vol. 24 No. 6
unions. As part of this argument, I pay attention to the effect of (1) the
degree of economic openness of a country to the world (economic link-
ages) as well as (2) the degree of linkages to the world system, specifically
the expansion of international discourse and the proliferation of a ‘scien-
tific consciousness’ (also known as institutional linkages). In addition, I
control for the effect of previous ICSU union memberships of a country
on its joining rate.
The findings of my dynamic analyses of historical data for 166 countries
between 1919 and 1990 are clear-cut. Institutional factors have the greatest
power to predict the rate at which nation-states join scientific organiza-
tions. The joining rate increases more quickly during the post-Second
World War era with the rise of the world system. Additionally, as the
number of ICSU union memberships of a country increases over time, the
hazard rate increases at a decreasing rate. By contrast, modernization
arguments do not seem to explain the evolution of ICSU membership.
These purely functional arguments are only important early in the ‘sci-
ence diffusion’ process. After 1945, institutional factors better account for
worldwide national scientific activity. Although core countries have
higher joining rates than peripheral countries, I find evidence of conver-
gence in the evolution of national joining rates over time. My longitudinal
results also demonstrate substantial invariability in the impact of func-
tional and institutional factors across countries, regardless of their level of
economic development. I conclude this article by discussing the implica-
tions of my research for the future study of national scientific production
and development in the world.
836
Castilla The Institutional Production of National Science
837
International Sociology Vol. 24 No. 6
this being especially so since these linkages represent key sources of pres-
sures for nation-states (from the world system) to behave ‘properly’ when
it comes to justifying the production of scientific knowledge.
In this line of thought, the central argument of institutionalists is that
the nation-state system is given worldwide support and legitimacy.
Interests and concerns of the state are increasingly determined by the
cultural content of the world polity (i.e. world definitions of the justifica-
tions, perspectives, purposes, aims and policies to be pursued by nation-
states). Thus, since 1945, we have witnessed an increase in the state’s
responsibilities and obligations along with a decrease in internal sover-
eignty. Both processes take place simultaneously. On the one hand, the
world system confers increasing powers on states ‘to control and organize
societies politically around the values established in the world political
culture’ (Meyer, 1980: 110). These powers are ideological, cultural and
political reflections of the logic of the world economy. At the same time,
modern nation-states’ organizations should be studied as similar social
actors under exogenous universal cultural processes. Increasingly, coun-
tries are structurally similar and change isomorphically. As world condi-
tions have changed, states have in many respects tended to become more
similar despite their internal differences (Grew, 1984).2 Thus, nation-states
do many similar things, such as ensuring mass education, respecting
human rights, taking environmental actions, etc. Indeed, states are
expected to be responsible for those actions. Nation-states are driven by
pressures and incentives, as well as changes in the world system. They
expand and become more similar in the functions they perform.
In sum, what all of these structures of the world polity produce is a
great deal of powerful talk rather than binding authoritative action, i.e.
scientific talk, legal talk, non-binding legislation, normative talk, talk
about social problems, suggestions, advice, consulting talk and so on
(Brusson, 1989, cited in Meyer, 1994: 9).3 Then, nation-states are supposed
to put into action the policies proposed in the international talk. Ultimately,
the main theoretical proposition in this article is that the rate at which
countries promote scientific activities should increase over historical time
with the increase in the dimensions of the world system. Here it is also
worth noting the prediction that as more countries, whether hegemonic or
peripheral, engage in the production of science and scientific activities,
other countries would tend to imitate such behavior in order to legitimate
themselves. As a result, national science and national scientific production
are more the result of an exogenous, worldwide, rationalistic culture that
defines ‘proper’ nation-state behavior when it comes to science – rather
than the result of nation-states as isolated actors selfishly evolving indepen-
dently of other nation-states (see Meyer, 2007).
838
Castilla The Institutional Production of National Science
Hypotheses
The existence and functioning of the ICSU provides an interesting object
of analysis because the process by which countries join any of its scientific
839
International Sociology Vol. 24 No. 6
700
600
500
400
300
200
100
0
1919 1929 1939 1955 1965 1975 1985
Year
840
Castilla The Institutional Production of National Science
Hypothesis 1b: The effect of national development on the joining rate of scien-
tific unions should increase over time.
Hypothesis 1c: More developed countries (core countries) are more likely to join
scientific unions than peripheral countries.
Having said that, a certain convergence in the evolution of joining rates
over time should also be expected. So that the difference in the likelihood
at which developed vs developing countries join ICSU scientific unions
should decrease over time.
Institutional Arguments
Following institutional approaches (as reviewed earlier), as the world
system rose and expanded after 1945, nation-states tended to behave iso-
morphically. In 1945, the United Nations was created and was thought to
legitimate nation-states and transmit the international rules and recom-
mendations concerning appropriate state structures and actions. Thus, it
could be predicted that:
Hypothesis 2a: The rate of joining ICSU scientific unions increases more quickly
during the post-Second World War era.
In addition to the historical time component, the linkage to the
expanded world society should also have an influence on national scien-
tific production. The hypothesis is then:
Hypothesis 2b: The degree of linkage of a country to the world system has a
positive effect on the rate at which a country joins a scientific union.
In order to measure the degree of national linkage to the world society,
I use two indicators. The first indicator is the cumulative number of
national memberships in ICSU unions. So that as the number of ICSU
union memberships of a country increases over time, the joining rate also
increases. The second indicator refers to the degree of economic linkages
to the world system as measured by the amount of imports and exports
as a percentage of GNP. Finally, and following the institutional predic-
tions, one should expect an inverted U-shaped relationship between the
rate at which countries enter scientific unions and the cumulative number
of union memberships of a country. Thus, the following is predicted:
Hypothesis 2c: As the number of ICSU scientific union memberships of a coun-
try increases over time, the rate at which a country joins a scientific union
increases at a decreasing rate.
Additionally, what all of these structures of the world polity produce is
some powerful talk. Nation-states are therefore supposed to (even pres-
sured to) put into action the policies proposed in international talks. So
the hypothesis here is the following:
841
International Sociology Vol. 24 No. 6
Hypothesis 3a: The rate at which a country joins a scientific union increases over
time with the increase in the dimensions of the world system.
In order to measure the dimensions of the world system, I use the num-
ber of international governmental organizations in the world, a standard
useful measure of the expanded world system. The reason for hypothesis
3a is that the international talks on science and development produced by
INGOs and the increasing involvement of nation-states in such talks
should increase the likelihood of a country entering any of the ICSU sci-
entific unions. Another related hypothesis is the following:
Hypothesis 3b: As more countries, whether core or peripheral, join scientific
unions, other countries imitate the behavior in order to become legitimate.5
Description of Data
The purpose of this study is to examine nation-state adherence to any
ICSU scientific union over the 1919–90 period. Specifically, I examine the
effects of both development and institutional processes on the rate at
which a country joins any union by estimating event history models. The
unit of analysis for the study is the nation-state and the event of joining a
union is a recurring event. The dimension of time used in the study is
historical time (i.e. 1919–90). At any given time, a nation may or may not
join any of the ICSU scientific unions.
The unique historical dataset analysed in this article is mainly obtained
from a larger cross-national cross-time dataset. This set provides informa-
tion for the period 1815–1990 by single years and by each country. It con-
tains over 250 variables with data on approximately 173 nation-states and
colonies. Primary sources include Arthur S. Banks’ Cross-National Time
Series Data Archive (Banks, 1986), Gurr’s Polity Persistence and Change Data
(Gurr and Gurr, 1978), The World Bank World Tables (World Bank, 1990), the
World Handbook of Social and Political Indicators (Taylor and Jodice, 1982),
various editions of the Yearbook of International Associations (Union of
International Associations, 1907–90) and several UNESCO Statistical
Yearbooks (UNESCO, 1990).
The sample analysed here includes 166 countries, after deleting cases
with missing information on the dependent variable. The data file con-
tains multiple records for each of the 166 nation-states – specifically, one
for each year. Since the event (i.e. a country’s joining of any ICSU scientific
union) is repeatable over time, information is available for a total of 6729
years of data for 166 countries (each of the countries is taken as multiple
observations over time). The data begin at the date of independence of
each country, or, in the case of old countries, in 1919 when the International
842
Castilla The Institutional Production of National Science
Research Council (the predecessor of the ICSU) was created. The data col-
lection ends in 1990, the final point for which data were available. There
are 713 events in total, 222 involving core countries and 491 involving
peripheral countries. The sample includes 21 core countries and 145
peripheral countries.
The variables were selected on the basis of: (1) relevance to the descrip-
tion of the major factors considered in this study; (2) judgment related to
the degree of homogeneity sufficient to allow cross-country comparisons;
and (3) availability of time series observations in most of the countries. A
detailed description of each variable is provided in the next subsection.
843
International Sociology Vol. 24 No. 6
18
16
14
12
10
0
1919 1929 1939 1955 1965 1975 1985
Year
Figure 2 Average Number of ICSU Union Memberships per Country per Year by
Different Categories of Countries (1919–90)
844
Castilla The Institutional Production of National Science
Results
Before proceeding to the multivariate methods of event history analysis,
it is crucial to undertake exploratory analysis of the longitudinal data
(Castilla, 2007). A better understanding of the temporal nature of the data
was obtained by estimating integrated hazard rates and hazard rates at
which countries join scientific unions over time. These functions were
estimated for the full sample of countries over the 1919–90 period. They
were also estimated for two subgroups of countries (i.e. core and periph-
eral) depending on the economic position of a country in the world.
The integrated or cumulative hazard rate indicates whether the hazard
rate varies with time. Thus, if the hazard rate is constant over time, the
integral of this rate over time should yield a straight line. Figure 3 pre
sents the integrated hazard rates for core and peripheral countries. The
non-linear relationship between the Nelson–Aalen estimates of the inte-
grated hazard rate at which nation-states join any of the ICSU scientific
unions and historical time suggests that the hazard rate is time depen-
dent, and therefore a constant (exponential) hazard rate model may not be
appropriate for the purposes of the multivariate analysis. Since the curve
bends upward, then the likelihood of a country joining at least one of the
unions increases over time. As the figure shows, the hazard rate is rela-
tively constant until around 1950. During the postwar era, the hazard rate
varies greatly over time. The plot partially supports hypothesis 2a in the
sense that the joining rate increases more quickly during the postwar era.
This is not surprising given that the UN, created in 1945, is believed to
have legitimated nation-states and transmitted the ‘blueprints’ concern-
ing nation-state action.
The slope of the integrated hazard rate is not constant for either group
of countries; however, it varies much more for core countries than for
peripheral countries. The slope of the integrated hazard rate for periph-
eral countries is approximately linear, especially prior to 1950. Although
the hazard rate for these countries may not be time dependent, the low
level of fluctuation may be in part due to the relatively small number of
peripheral countries joining the ICSU unions, especially prior to 1945. The
plot clearly indicates that the rate of adoption throughout the entire
period is higher for core countries.
More information is gained from examining the hazard rate itself. The
hazard rate is similar to the probability that a nation-state joins any of the
ICSU scientific unions per unit of time. Figure 4 shows the smoothed
Nelson–Aalen estimates of the hazard rates for core and peripheral
845
International Sociology Vol. 24 No. 6
10
8
6
R(t)
4
2
0
Figure 3 Estimates of the Integrated Hazard Rate at Which Nation-States Join ICSU
Scientific Unions over Historical Time by Category of Countries (1919–90)
Note: Number of events/countries: 222/21 (core countries); 491/145 (peripheral countries).
countries over the 1919–90 period.10 It confirms that the hazard rate for all
countries changes non-monotonically over time. After smoothing, the
hazard rate remains quite constant until the 1945 period, when it rises
sharply to more than half this rate again over the following 10 years (from
1950–60).11 The post-1960 period is characterized by a slight decline in
the rate; this decline continues after 1970, even when new countries
became independent and when seven new ICSU unions were created
(two in the 1970s, two more in 1982 and three more in 1993).12
Institutional arguments seem to apply on a world level to the joining of
ICSU scientific unions, as theory predicts that the highest rates of joining
will occur with the rise of the world system in the postwar era. Thus, the
formation of the UN in 1945 and the founding of UNESCO soon after
contributed to the sharp rise in the hazard rate after 1950. A possible func-
tional explanation of the burst of activity culminating in 1960 could be
that the immediate postwar period was characterized by economic expan-
sion and growth. The time-variant nature of the gross national income per
capita used in this analysis will help in assessing this suggestion later in
the multivariate analyses.
When examining the differences between the two smoothed hazard rates
for the two different categories of countries, it is clear that their overall
846
Castilla The Institutional Production of National Science
.25
.2
Smoothed r(t)
.15
.1
.05
0
Historical year
Figure 4 Smoothed Estimates of the Hazard Rate at Which Nation-States Join ICSU
Scientific Unions over Historical Time by Category of Countries (1919–90)
Note: Number of events/countries: 222/21 (core countries); 491/145 (peripheral countries).
847
International Sociology Vol. 24 No. 6
memberships is 3.75 for core countries and 0.4 for peripheral countries. The
average number of memberships increases at a rate of 2.23 per decade in
the case of core countries (vs a rate of 0.6 per decade in the case of periph-
eral countries). Since the curves for the two different categories of countries
are not equally spaced within the 1919–90 observational period, the hazard
rates for core and peripheral countries are not multiples of one another
and, therefore, proportional hazard rate models do not seem appropriate
for modeling the phenomena under study.14 Moreover, the curves for the
two groups differ quite dramatically. The differences are greatest prior to
1960, when the joining activity by peripheral countries is very low. The
lines only seem to approach each other by the end of the period analyzed,
with the rise and expansion of the worldwide polity system.
848
Table 1 Piecewise Models of the Hazard Rate at Which Nation-States Join ICSI Scientific Unions in Three Historical Periods (1919–45,
1946–64 and 1965–90)
Covariates 1919–45 1946–64 1965–90 1919–45 1946–64 1965–90 1919–45 1946–64 1965–90
1. Time-independent
vector of covariates (b)
Intercept -1.759**** 0.026 3.538**** -7.409**** -30.250**** -35.450**** -22.900**** -38.210**** -53.640****
Gross national income 0.626* 0.248** 0.016 1.044** 0.267** 0.063**** -1.653* 0.799 -0.686**
per capita (in 000s)
Imports and exports -0.002 -0.001 0.000 -0.002 0.002 0.000 0.000 0.001 0.000
(as % of GNP)
Cumulative number of 0.594**** 0.319**** 0.386**** 0.567**** 0.382**** 0.397**** 0.642**** 0.355**** 0.395****
national memberships
in ICSU scientific
unions
Squared cumulative -0.072**** -0.017**** -0.017**** -0.071**** -0.020**** -0.018**** -0.079**** -0.018**** -0.018****
number of national
memberships in ICSU
scientific unions
Cumulative number of -0.035**** -0.038**** -0.017**** -0.044**** -0.090**** -0.034**** -0.051**** -0.099**** -0.032****
ICSU union
memberships in
existence in the world
Number of INGOs in 0.772*** 0.369**** 0.037**** 3.210**** 0.163**** -0.158* 4.134**** 4.794 **** -0.045*
the world (in 00s)
Core countries (dummy 0.581** -0.082 -0.052 0.582** -0.037** -0.012 0.597 -0.055 0.420
variable)
(Continued)
Table 1 (Continued)
Covariates 1919–45 1946–64 1965–90 1919–45 1946–64 1965–90 1919–45 1946–64 1965–90
2. Time-dependent vector
of covariates (W)
Intercept -0.371**** 1.346**** 1.157**** 4.739**** 1.598**** 1.497****
Core countries (dummy -0.005 0.006 -0.007
variable)
Gross national income 0.245** -0.013 0.011**
per capita (in 000s)
Number of INGOs in -0.079**** -0.007**** -0.002****
the world (in 000s)
Overall rate per year 0.073 0.325 0.193 0.073 0.325 0.193 0.073 0.325 0.193
Number of events 71 270 372 71 270 372 71 270 372
Chi-square statistic
Relative to the baseline
modela 588**** 1057**** 1210****
Degrees of freedom 21 21 30
Relative to the piecewise
exponential model 470**** 622****
Degrees of freedom 3 12
Relative to the
piecewise proportional 153****
Gompertz model
Degrees of freedom 9
Number of countries = 166
* Significant at the .10 level; ** significant at the .05 level; *** significant at the .01 level; **** significant at the 0.001 level.
a
The model labeled baseline is a model with no covariates (constants only).
Castilla The Institutional Production of National Science
851
International Sociology Vol. 24 No. 6
μ = Ω’Z
where Z is a vector of covariates which may or may not overlap with the
vector of covariates X. Essentially, the variation in the hazard rate over
time depends now on the vector of covariates Z. The functional form for
the hazard rate is now:
852
Castilla The Institutional Production of National Science
Table 2 Piecewise Gompertz Model of the Hazard Rate at Which Core Countries Join
ICSU Scientific Unions in Three Historical Periods (1919–45, 1946–64 and
1965–90)
Core countries
* Significant at the .10 level; ** significant at the .05 level; *** significant at the .01 level.
a
The model labeled baseline is a model with no covariates (constants only).
853
International Sociology Vol. 24 No. 6
Table 3 Piecewise Gompertz Model of the Hazard Rate at Which Peripheral Countries
Join ICSU Scientific Unions in Three Historical Periods (1919–45, 1946–64
and 1965–90)
Peripheral countries
* Significant at the .10 level; ** significant at the .05 level; *** significant at the .01 level.
a
The model labeled baseline is a model with no covariates (constants only).
the data significantly better than the other two. In order to further exam-
ine whether the effects of the covariates differ by category of country,
I also estimate two piecewise non-proportional models for core and peri
pheral countries (reported in Tables 2 and 3). Results are discussed in the
following section.
854
Castilla The Institutional Production of National Science
Discussion of Results
Table 1 presents several piecewise hazard rate models estimated for all
countries in three historical periods from 1919 until 1990. Columns 7–9
report the maximum likelihood estimates of the coefficients for the piece-
wise non-proportional Gompertz model of the hazard rate at which coun-
tries join at least one of the ICSU unions.21 The model does not provide
support for hypothesis 1a that socioeconomic development has a strong,
positive effect on the ICSU union joining rate. On the contrary, the sign of
the coefficients related to the GNP per capita measure is negative in the
1919–45 and the 1965–90 periods (only significant at the .05 level in the
latter period). This suggests that the relationship between development
and joining rates should not be hypothesized in purely internal functional
terms. In addition, the effect of GNP per capita appears to have an insig-
nificant positive effect (p < .05) on the rate during the 1946–64 period.
Nevertheless, by looking at the effects of GNP on the hazard rate as
estimated by the piecewise proportional Gompertz model, one can see
that socioeconomic development has a strong positive effect on the rate at
which countries join scientific unions early in the period of analysis. Thus,
a US$1000 increase in GNP per capita increases the joining rate by 184
percent during 1919–45 (p < .05), ceteris paribus.22 The same GNP increase
is estimated to increase the rate by 30 percent in the following period
(p < .05). During the 1965–90 period, a US$1000 increase in GNP per capita
increases the hazard rate by 6 percent (p < .001). These results do not sup-
port hypothesis 1b in the sense that the effect of socioeconomic develop-
ment on the joining rate is not found to increase during the three historical
subperiods. Instead, the results suggest the opposite trend. Thus, devel-
opment arguments might predict the national process of joining any sci-
entific union early in the ‘science diffusion’ process but not after the
process is well under way (consistent with Tolbert and Zucker, 1983).
More remarkable results are found to support institutional arguments
in this study. The cumulative number of national memberships in ICSU
scientific unions, which reflects the extent to which a nation-state is linked
to the world system and its discourse on science, has a strong significant
positive effect on the hazard rate at which nation-states join ICSU unions
(p < .001). Thus, in the 1965–90 period, having one more membership in
any of the scientific unions increases the joining rate by almost 50 percent.
Moreover, this degree of national linkage of a country to the world-scien-
tific system increases the hazard rate at a decreasing rate. This is shown
by the negative sign associated with the squared cumulative number of
national memberships in scientific unions.23 In sum, there is evidence to
support hypotheses 2b and 2c in the sense that there is an inverted
U-shaped relationship between the rate at which countries join ICSU
855
International Sociology Vol. 24 No. 6
856
Castilla The Institutional Production of National Science
857
International Sociology Vol. 24 No. 6
the first to join. Once the majority of core countries have joined the exist-
ing scientific unions by 1965, peripheral countries follow and copy previ-
ous core-nation scientific behavior.
Conclusion
In the literature on the development of science and world polity, signifi-
cant progress has been made in demonstrating the conditions under
which national science and scientific production promote economic and
social development. However, less understood are the underlying factors
that have influenced the emergence of national scientific activity around
the world, especially during the 20th century. For the first time in this
literature, this article presents unique cross-national, cross-time empirical
research on the factors influencing national scientific activity. In particu-
lar, I examine in-depth the key factors affecting the rate at which countries
join any of the unions comprising the ICSU, the pre-eminent and oldest
international science institution in the world, over the 1919–90 period.
According to the theoretical framework proposed here, institutional
theories have greater power to predict the rate at which nation-states join
scientific unions over time. Thus, the joining rate increases more quickly
with the rise of the world system during the post-Second World War era.
In addition, the findings in this study suggest that as a country’s ICSU
union memberships increase over time, the hazard rate at which a nation-
state joins an additional ICSU union increases at a decreasing rate. This
inverted U-shaped relationship between the rate and the cumulative
number of a country’s union memberships holds for both core and
peripheral countries. Furthermore, the increasing dimensions of the
world system as well as the impact of its discourse on nation-state behav-
ior are also proven to have a strong positive effect on the hazard rate for
both core and peripheral countries.
Development and modernization arguments do not seem to offer sig-
nificant explanations for the ICSU joining hazard rate over time. The
results have shown that more developed countries are more likely to join
ICSU unions than peripheral ones. However, the estimated longitudinal
models do not provide support for the hypothesis that socioeconomic
development has a strong positive effect on the rate at which countries
join ICSU scientific unions. Instead, they prove that development argu-
ments explaining the national process of joining any ICSU union might
predict joining early in the ‘science diffusion’ process (before 1965). After
1945, institutional factors better account for the nature and change of a
nation’s rate of joining ICSU scientific unions, as predicted by the neo
institutional theoretical arguments.
858
Castilla The Institutional Production of National Science
Finally, the results confirm that core countries have significantly higher
hazard rates than peripheral countries. There is also evidence of conver-
gence in the evolution of joining rates over time. Thus, the difference in
the likelihood at which developed vs developing countries join ICSU
unions seems to have decreased slowly over time. The results in this
article demonstrate that there is no substantial variability in the effects of
the different development and institutional covariates for core and
peripheral countries. This finding confirms that any country in the world
today (regardless of its welfare, culture or history) can experience the
institutional pressures to act rationally and reasonably in the interest of
activities meant to develop science and increase scientific production.
The findings of this study have important implications for the future
study of national scientific activity and development in the world. First,
my cross-national, cross-time empirical research on the factors influenc-
ing national scientific activity confirms that scientific activities have
become a defining rationalizing force in any national modern system
(consistent with Drori et al., 2003). In 1931, 38 nation members had 193
different memberships in the different ICSU scientific unions; the joining
of countries to ICSU unions began early in the century, and steadily
increased during the postwar era (see Figure 1). Second, my study sup-
ports a more institutional account for the process of national scientific
production over time. It has been quite common to stress the expansion of
science as instrumental to the self-interest of countries, particularly to the
specific goals of certain social groups. However, by applying the event
history methodology to analyze repeating national-level joining events
since 1919, I show that institutional arguments better account for the
expansion of national scientific memberships as well as other scientific
activities; this is so even after controlling for key modernization argu-
ments in the analyses. Last but not least, this article expands the scope of
previous empirical studies by analyzing unique data on how and when
countries join any of the scientific unions of the ICSU.
As with the results of any study based on one sample drawn from a
single international science organization, caution needs to be exercised
when generalizing the results of this study. Thus, I believe that this
research can be extended in several interesting theoretical and empirical
directions. The first and most obvious area to be explored involves devel-
oping studies to continue testing the institutional aspects of national sci-
entific activities as well as science production in more comprehensive and
detailed ways. I have explored when and under what conditions coun-
tries join any of the scientific bodies comprising the ICSU. I focus on ICSU
unions as a matter of necessity. In particular, one can believe in the insti-
tutional aspects of national scientific production without believing that it
859
International Sociology Vol. 24 No. 6
has to come through joining ICSU unions. After all, once a country has
joined any of the ICSU unions as well as many other international organi-
zations, other institutional pressures may exist to develop scientific pro-
duction at the national level. Studies need to look at the institutional
arguments presented here by including dynamic information about coun-
tries’ multiple scientific activities and enterprises before and after they
join any international scientific organization. It is also important to collect
detailed information about a greater range of national-level scientific
actions over time. Additional research dealing with multiple scientific ties
to other scientific institutions (and their effect on national development)
is much needed.
The second extension is closely related to the first. Although previous
studies have documented relevant institutional similarities in a wide vari-
ety of national-level structures and institutions, most of them exclusively
focus on very specific national measures of activities and outcomes. In my
current study, I only look at how the degree of linkage of a country to the
world system affects its joining rates of ICSU scientific unions. One point
often emphasized in the neoinstitutional literature, however, is that the
institutional factors that affect national-level processes and outcomes are
decidedly much more complex than the available measurement proxies
used to account for their influence at the national level of analysis. In this
regard, more research is required to understand how other sources of
influence, such as expansion of universities and their enrollment, govern-
mental economic institutions, private sector enterprises, as well as the
globalization of practices and routines may affect the expansion of science
at the national level. More studies are also needed not only on how these
factors influence scientific production but also how they ultimately pro-
mote countries’ economic and social development. Without such compre-
hensive studies that collect and analyze data on more fine-grained
national-level measurements of scientific activities, science production
and linkage to the world community, many of the intervening mecha-
nisms are still left open to speculation.
In sum, my examination of the evolution of the ICSU scientific member-
ship suggests the necessity of cross-national comparative research for a
more comprehensive and thorough empirical evaluation of national insti-
tutions and scientific production over time. Social scientists interested in
comparative analyses on national scientific production need more and
better data on national structures and institutions that could account for
differences across countries. Science and scientific production should no
longer be reduced to the interaction of economic and modernization fac-
tors; they should be regarded as sociological processes of interest to those
curious about the impact of science on national development around the
world. In particular, nation-states are influenced (even constructed) by a
860
Castilla The Institutional Production of National Science
861
International Sociology Vol. 24 No. 6
level. Such analyses are indispensable tools for understanding the circum-
stances and institutional factors that promote successful economic and
social development outcomes.
Notes
An earlier version of this article was presented at the Annual Meeting of the
American Sociological Association and in Castilla (2007). I am indebted to John W.
Meyer and Mark Granovetter for their guidance and support. I have benefited
greatly from the comments and suggestions of Hokyu Hwang, Francisco Ramirez,
Evan Schofer, Nancy B. Tuma, Kiyoteru Tsutsui and the three International Sociology
anonymous reviewers. I also wish to thank the members of the Stanford Comparative
Workshop for their helpful advice and comments on earlier drafts of this article.
1. Once again, several studies analyze the impact of world institutional models
on national systems in a variety of domains (see the examples listed at the
end of the first paragraph in this article).
2. Activities and operations undertaken by certain leading economies are pro-
gressively copied and mastered by other economies. Thus, the US, Canada
and Japan have provided models to be copied by other countries throughout
the century. However, in the modern period, Strang and Meyer (1993) claim
that countries are less likely to copy successes directly than to copy them as
institutionalized and interpreted by the world system.
3. An example are the world’s scientific communities, which produce a power-
ful talk about environmental regulations, about failures and requirements in
national development policies and about the protection of human rights.
4. According to its statutes, the principal objectives of the ICSU are: (1) to
encourage and promote international scientific and technological activity for
the benefit and well-being of humanity; (2) to facilitate coordination of the
activities of the scientific union members (see statute 7); (3) to facilitate coor-
dination of the international scientific activities of its national scientific mem-
bers (see statute 8); (4) to stimulate, design, coordinate and participate in the
implementation of international interdisciplinary scientific programs; (5) to
act as a consultative body on scientific issues that have an international
dimension; and (6) to engage in any related activities.
5. In other words, the rate at which countries become members of ICSU scien-
tific unions should increase with the rise in the cumulative number of ICSU
scientific union memberships in existence in the world over time.
6. Data were available on the year when a country joined an ICSU union. It was
impossible to know the exact date when the event took place though (i.e.
month and day). Additionally, the fact that a country may join one or more
unions in a given year should not be seen as different unrelated events.
Instead they could be taken as one single event. This is because it is hard to
believe that the processes of a country joining one or more scientific unions
in a given year are completely independent from each other, and therefore the
862
Castilla The Institutional Production of National Science
same explanatory factors may account for the ‘effort’ of joining one, two or
more unions in a given year. Nevertheless, I control for the number of ICSU
memberships of a given country as an independent variable in the analyses.
7. Less than 1 percent of the earlier values of GNP are extrapolated using the
linear extrapolation facility available in SPSS for Windows (i.e. I use the method
of imputing the missing values with estimated values using regression meth-
ods). The same results are obtained when using other standard measures of
development (e.g. the iron/steel production for the early period).
8. The ICSU, as the primary international scientific institution, has been argued
to contribute to the creation and maintenance of a worldwide discourse and
normative pressure to promote science (see Thomas et al., 1993).
9. The pattern of membership growth for core and peripheral countries was the
opposite before 1960. Thus, in 1920 the average number of memberships was
3.75 for core countries (20 core countries) and 0.4 for peripheral countries (36
peripheral countries). The average number of memberships increases at a rate
of two additional memberships per year in the case of core countries vs a rate
of almost six in the case of peripheral countries. There is also a trend toward
convergence in the average number of national memberships per country.
10. I also estimated the unsmoothed Nelson–Aalen estimates of the hazard rate
at which countries join any of the ICSU scientific unions for the full sample of
countries (available upon request). Spurts of joining appeared between 1925
and 1930, later between 1950 and 1977, and finally after 1982. The rate peaks
around 1950, when about 51 percent of all the nation-states at risk (39 out of
76 countries) joined at least one ICSU union. The hazard rate for the second
highest peak occurs in 1962 (50 percent of all 111 countries at risk joined at
least one scientific union). Finally, the third highest peak occurs in 1975, when
about 40 percent of the nation-states at risk (61 out of 142 countries) adhere to
at least one union.
11. The smoothed hazard rates are considerably smaller than the unsmoothed
hazard rates at corresponding historical times, as smoothing levels out the
‘erratic’ tendencies of the hazard rate.
12. In 1993, three new ICSU unions were created: the International Society of Soil
Science (ISSS), the International Brain Research Organization (IBRO) and the
International Union of Anthropological and Ethnological Sciences (IUAES).
Since this study only covers the 1919–90 period, these three unions are
excluded from the analyses. So the maximum number of scientific unions
analysed here is 21.
13. Before 1960, 225 events take place. Forty-eight percent of those events take
place in core countries (21 core countries) and the rest take place in 43 differ-
ent peripheral countries (115 events). The opposite occurs during the post-
1960 period. Thus, after 1960, 488 events took place. Seventy-seven percent of
those events occur in 91 peripheral countries. Eighty-two percent of those
events are joinings by peripheral countries which became independent before
1960 (59 countries). Thirty countries that became independent after 1960 only
account for 69 events in the 1960–90 period.
14. For more information, see Castilla (2007) and Tuma and Hannan (1984).
863
International Sociology Vol. 24 No. 6
864
Castilla The Institutional Production of National Science
22. The percentage change in the hazard rate associated with a unit change in
GNP per capita is 184.06 percent = (exp(1.044) − 1) × 100, ceteris paribus.
23. To test hypotheses 2b and 2c, I include the cumulative number of national
memberships in ICSU scientific unions both in linear and quadratic form
in the models. Hypothesis 2b is not supported when I use the degree of
economic linkages to the world system indicator. Thus, the amount of
imports and exports as a percentage of GNP has neither a significant
impact nor the predicted positive effect on the hazard rate of joining any
scientific union.
24. 62.42 = exp(4.134) and 120.78 = exp(4.794).
25. For purposes of simplicity, these particular results are not presented here, but
they are available upon request.
26. I was able to take this approach because I have a relatively large sample
of events for core and peripheral countries (number of events/countries):
222/21 (core countries) and 491/145 (peripheral countries). Event datasets
on countries are not typically large enough to support extensive analysis of
this type.
27. In preliminary analyses, I estimated a set of hazard rate models with interac-
tion effects between all the model covariates and the dummy variable for
category of country (i.e. core vs peripheral country); I always found results
similar to the ones I report in this article. These analyses are available upon
request.
28. 1.19 = exp(0.171) and 1.46 = exp(0.377).
References
Banks, A. S. (1986) Cross-National Time-Series Data Archive 1815–1986 (CNTS).
Binghamton, NY: Center for Comparative Political Research.
Baum, R. C. (1974) ‘Beyond Convergence: Toward Theoretical Relevance in
Quantitative Modernization Research’, Sociological Inquiry 44: 225–40.
Berkovitch, N. (1993) ‘From Motherhood to Citizenship: The World-Wide
Incorporation of Women into the Public Sphere in the Twentieth Century’, PhD
dissertation, Stanford University, CA.
Blossfeld, H. P and Rohwer, G. (1995) Techniques of Event History Modeling: New
Approaches to Causal Analysis. Mahwah, NJ: Lawrence Erlbaum.
Boli, J. (1987) ‘World Polity Sources of Expanding State Authority and Organizations,
1870–1970’, in G. M. Thomas, J. W. Meyer, F. O. Ramirez and J. Boli (eds)
Institutional Structure: Constituting State, Society and the Individual, pp. 71–91.
Beverly Hills, CA: Sage.
Boli, J. (2005) ‘Contemporary Developments in World Culture’, International
Journal of Comparative Sociology 46: 383–404.
Castilla, E. J. (2004) ‘Organizing Health Care: A Comparative Analysis of National
Institutions and Inequality over Time’, International Sociology 19(4): 403–35.
Castilla, E. J. (2007) Dynamic Analysis in the Social Sciences. San Diego, CA:
Elsevier/Academic Press.
Charles, M. (1992) ‘Cross-National Variation in Occupational Sex Segregation’,
American Sociological Review 57: 483–502.
865
International Sociology Vol. 24 No. 6
866
Castilla The Institutional Production of National Science
867
International Sociology Vol. 24 No. 6
868
Castilla The Institutional Production of National Science
869