Extending the INGO Network Country Score , 1950 – 2008

Hughes et al. (2009) introduced the INGO Network Country Score (INCS), a measure of country-level connectedness to the world polity, for three years: 1978, 1988, and 1998. The measure scores countries by centrality in the world country-INGO network, rather than on raw counts of INGO ties that do not acknowledge networks or power. In this article, we extend the measure by time, space, organization, and calculation. First, we extend the measure to the period 1950–2008, allowing closer correspondence to the years typically assessed by researchers. Second, we extend the country samples upon which the scores are based, allowing researchers greater flexibility in their analyses. Third, we extend the number of INGOs from which the scores are created. The Hughes et al. (2009) INCS were based on a single-year maximum of 476 INGOs; ours are based on a single-year maximum of 1,604 INGOs (5,291 INGOs across all years). Finally, we provide both raw and scaled scores. We discuss the observed increasing density in the world polity from 1950 to 2008, comparing scores across regions. Results reveal higher average INCS with less variability among Western countries, and significant inequality between the West and the rest of the world.

B ASED largely on the impressive growth in international non-governmental organizations (INGOs) over the last century, scholars claim that states are increasingly connected to each other through their INGO ties (e.g., Boli andThomas 1997, 1999;Schofer 2003).Thousands of INGOs, ranging from scientific associations to sports groups, have been founded since the mid-nineteenth century.The growth of these organizations over time is seen as fostering connections between states and ultimately leading to a dense global network of states and organizations-the "world polity" (Boli andThomas 1997, 1999;Meyer et al. 1997).
In a world "increasingly integrated by networks" (Boli and Thomas 1997:172), countries embedded in this global network are more likely to adopt world models and standards of organization and behavior (Meyer et al. 1997).Indeed, empirical studies link INGOs to a range of outcomes, including education funding (e.g., Kim and Boyle 2012), environmental protection legislation (e.g., Frank, Hironaka, and Schofer 2000;Jorgenson, Dick, and Shandra 2011), women's rights (Ramirez, Soysal, and Shanahan 1997;Paxton, Hughes, and Green 2006), and improved human rights practices (Boyle and Kim 2009;Cole 2005).World polity theory is therefore fundamentally concerned with a global network structure hypothesized to act as a social force on nation-states.
Typically, connectedness to the world polity is measured using counts of INGO memberships.INGOs are therefore treated as a country attribute-an attribute acquired by a country when its citizens join INGOs-similar to other state attributes such as GDP or democracy.But the central argument of Hughes et al. (2009) is that INGOs are more than a state attribute.As world polity theory stresses, INGOs also create networks that define the relative positions and power of nations in the international system.Rather than thinking of INGO counts as measuring a state's ties to the world polity, Hughes et al. argue that we must consider the structural position that a state occupies in the larger network of INGOs.Their argument was similar to those of scholars who have considered how states are positioned in networks of intergovernmental organizations (Beckfield 2010;Hafner-Burton and Montegomery 2006) or trade (Clark and Beckfield 2009;Ingram, Robinson, and Busch 2005;Grey and Potter 2012).Hughes et al. (2009) developed a new over-time measure of country-level connectedness to the world polity, the INGO Network Country Score (INCS).This measure scores countries by their centrality in the world country-INGO network.Hughes et al.'s key insight was that "two countries can have the same raw count of INGO memberships but occupy different positions in the overall network created by countryto-INGO ties.Having ties to countries (through joint membership in INGOs) that are more embedded in the network moves a country toward the center of the network, and likely closer to sites of both diffusion and influence." While this was certainly an important advance, there are several limitations to the INCS, all of which restrict researcher flexibility and therefore reduce the measure's usefulness.First, Hughes et al. (2009) present INCS for only three time points, spaced 10 years apart : 1978, 1988, and 1998.This is only a fraction of the years typically assessed by researchers in the area.Second, INCS are calculated for only two country samples (time-varying and constant), both of which exclude countries with fewer than 1 million residents.Third, the measure includes only universal membership organizations in calculating their scores; researchers may hesitate to use a score based on a small subset of the total population of INGOs.Finally, Hughes et al. (2009) provide only scaled centrality scores, even though unscaled scores could be desirable in some research designs.
In this research note, we update and extend the INCS to address these limitations.First, and most importantly, we extend the measure to cover approximately every five years over the period 1950 to 2008: 1950, 1955, 1965, 1973, 1978, 1983, 1988, 1993, 1998, 2003, and 2008.The vast majority of research on the influence of the world polity considers a range of dates within 1950-2008, so the updated INCS will have greater utility to a greater number of researchers.Second, as calculation of the network scores depends on the countries in the sample, we present the INCS for four separate country samples: all countries, all countries with populations over 1 million, a set of countries that have been in existence continually since 1955, and a set of countries that have been in existence continually since 1978.Together these samples should capture the sets of countries most commonly analyzed by cross-national researchers.Third, we significantly extend the number of INGOs used to calculate the network measures.In order to focus on global rather than regional organizations, Hughes et al. (2009) restricted their sample of INGOs to "universal membership organizations."However, "intercontinental organizations" also allow cross-regional ties.Here, we include such organizations, extending the number of INGOs used to calculate the network measures to 5,291 across all time sociological science | www.sociologicalscience.com periods (from 282 in 1950 to 1,604 in 2008).Finally, in addition to a normalized version of eigenvector centrality, where eigenvector centrality is divided by the maximum difference possible in a given year expressed as a percentage (Borgatti, Everett, and Freeman 2002), we present an alternative unscaled country score.The raw eigenvector scores could be used by researchers investigating global processes in a single year.Altogether, the updated and expanded INGO Network Country Score will be of much greater use to researchers of the world polity.
The 2009 INGO Network Country Score-Theoretical Improvement but Limited Utility Connectedness to the world polity is typically measured as an attribute of a country-as country-level counts of INGO memberships that stratify countries by their raw number of INGO memberships.But as Hughes et al. (2009) explained, a count measure is problematic because it assumes that all connections are equivalent-that all INGOs are equally able to link a country to the world polity.Raw counts tell us nothing about how states are tied together and do not adequately account for differences in power in the network (Beckfield 2003(Beckfield , 2010;;Hafner-Burton and Montegomery 2006).
The distinction matters because it is the INGO network, rather than any simple state attribute, that allows states to exchange information, support, and authority (Granovetter 1973;Friedken 1993;Lin 1999;Moody 2001;Bearman and Bruckner 2001;Ingram, Robinson, and Busch 2005;Haefner-Burton and Montegomery 2006).The overall network influences the behavior of states by increasing the power of some and by shaping shared norms about behavior.On the one hand, power is endowed by network position and allows for the promotion or alteration of collective beliefs.Much like people, states hold positions that are more or less prestigious in the social network.Thus, in the international system, some central states have a great deal of social power while others have less.On the other hand, pressure on states to conform to international conventions is likely to emerge from the environment created by the network of states.States in more central positions should be more amenable to common beliefs and norms held in the world polity and therefore more likely to adopt indicators of those beliefs, such as human rights norms.Position makes certain actions more compelling than others (Hafner-Burton and Montegomery 2006).
To address the limitations of count measures, Hughes et al. (2009) used social network analysis to determine the structural position of states in the network formed by INGOs.They measured embeddedness as a function of the shared ties that states have to each other through INGO membership, taking into account the pattern of ties in the entire network.They analyzed the affiliation network created by country-level membership within a set of specific INGOs (Breiger 1974;Wasserman, Faust, and Iacobucci 1994;Borgatti and Everett 1997;Faust et al. 2002).Each cell of the affiliation matrix is the sum of shared INGO ties that two countries have in common.They then calculated the network centrality, or "network positionconferred advantage" (Cook and Whitmeyer 1992:120), of each country, using the Bonacich (1987) eigenvector measure of network centrality.Eigenvectors, also called prestige scores, describe the extent to which countries are connected to other highly connected countries (Haefner-Burton and Montegomery 2006;Ingram, Robinson, and Busch 2005;Beckfield 2010; for broader applications, see also Mizruchi 1993;Podolny 1994;Cornwell and Harrison 2004).Eigenvector centrality allows for distinctions to be made between memberships that pull a country closer to the core of the network and those that do not.Countries with higher centrality scores are connected to many actors or a few central actors; countries with lower scores are not only less connected overall, but they are less connected to central countries.For more detail on the creation of the INCS, see Hughes et al. (2009).
The INCS measure was unquestionably a theoretical improvement over raw country counts, but the limitations of the INCS listed above lessen the likelihood of wide adoption by global and transnational scholars.In what follows, we introduce an updated set of INCS that avoids these problems and should be of greater utility to a wider range of scholars.

Data
Using the Yearbook of International Organizations, we recorded country-level memberships in 5,291 INGOs for 198 countries in eleven time periods : 1950, 1955, 1965, 1973, 1978, 1983, 1988, 1993, 1998, 2003, and 2008.We identified all INGOs with potential "universal" membership, excluding organizations that had an exclusive membership, organizations that were intergovernmental rather than non-governmental, organizations with a single regional focus, and organizations which were institutes, foundations, or centers.INGOs that crossed regional lines, such as "intercontinental" organizations, were included. 1 Because the eigenvector measure of network centrality, on which the INGO network country scores are based, is fully relational, the scores vary depending on which countries are included in the network.In other words, a single country's position in the network-and thus its score-is affected by every other country in the network.Thus selecting the sample of countries to include in the network and subsequent calculation is crucial.Each year of the dataset includes a different set of countries (countries are born, merge, and die).Using the maximum number of countries in a given year is likely the best strategy for a researcher interested in a cross-sectional analysis.But "constant" samples that exclude country births and deaths are important for certain longitudinal analyses.For example, in looking at network measures over time, we would like to know whether density and other measures are affected by the introduction of new countries, or whether they are changing over time regardless of the entry and exit of nations.
For these reasons, we created four samples of countries that cover a wide range of possibilities: (1) the full sample of sovereign countries in existence in any year; (2) a sample limited to sovereign countries with a population over 1 million citizens in 2000 (a restriction common to much cross-national research); (3) a constant sample of 78 countries that were independent throughout the period, from 1955 on; and (4) a constant sample of 126 countries that were independent from 1978 on.To reiterate, the 1978 constant sample excludes countries that were "born" between 1978and 2008(e.g., Azerbaijan) or "died" between 1978and 2008 (e.g., Yugoslavia). 2  Before calculating centrality, we also imputed data for a small subset of cases.First, we imputed country memberships for INGOs that were listed in the yearbook in a given year but for whom no members were listed. 3If an INGO appeared in the yearbook and had members in other years, then we considered it likely that its lack of members in the given year was not informative, and considered imputing country members to be reasonable.Second, we imputed country memberships for INGOs that did not appear in a single year's yearbook but did appear in either two adjacent prior yearbooks and one adjacent following yearbook, or the reverse (one adjacent prior yearbook and two adjacent following yearbooks).The second strategy results in imputation of a small, reasonable number of INGOs that did not appear in the book, but for which adjacent membership data suggests a reasonable expectation that the organization continued to exist and have members.There is a vanishingly small probability that an organization became inactive, and then reactivated within the span of a few years. 4 Imputation proceeded as follows.First, we created a variable that flagged a country if it was a member in the preceding or following year of available membership data for the INGO.The idea is that we only want to impute reasonable countries-those that were members of the INGO before or after the year in question.Second, we estimated the expected number of members for each organization at each point in time using a multi-level random intercept Poisson model based both on the overall number of members for all organizations and on the number of members for that organization in other years.The choice of random intercept allowed our estimate to vary by year and by organization.Third, we created a random prediction of which countries from the relevant pool of countries would be members of the INGO.For this we used ranks derived from randomized Bayesian predicted probabilities, where countries were ranked in order of decreasing probability of membership, and only countries with ranks lower than the expected number of members for the INGO were given a membership. 5 Overall, we imputed memberships representing approximately seven percent of the total observations with data, but only two percent of all possible observations.Scores based on unimputed data are similar and are available in the appendix.
Finally, we created affiliation networks for each of the four samples of countries in each of the eleven years.Each cell of the affiliation matrix is the sum of shared INGO ties that two countries have in common.We then calculated the unscaled and scaled network centrality of each country, using the Bonacich (1987) eigenvector measure of network centrality.

INGO Network Country Scores 1950-2008
Table 1 presents the scaled INCS for the full sample of countries with greater than 1 million population for six of the eleven collected time points : 1955, 1965, 1978, 1988, 1998, and 2008.The appendix presents scaled and unscaled INCS for all eleven years of data, and for all samples-the full sample of sovereign countries in existence in any year; the sample limited to sovereign countries in existence in any sociological science | www.sociologicalscience.com year that had a population over 1 million citizens in 2000; the constant sample of 78 countries that were independent throughout the period from 1955 on; and the constant sample of 126 countries that were independent from 1978 on. 6 1 shows interesting patterns of variation and change.Because network measures are explicitly relational, individual country numbers should be interpreted in relation to other countries.For example, Greece's 2008 score of 0.83 suggests that it is roughly 20 percent less connected than Germany, which is the most highly connected country in that year.It becomes apparent that although country network scores may fluctuate from year to year, broader inequalities in the network remain more constant.For example, Italy is always highly central in the network, and the Republic of the Congo is not.Expected patterns appear: for example, after the Korean War, South Korea becomes increasingly integrated into the world polity (0.15 in 1955 to 0.76 in 2008), while North Korea remains highly marginalized (0.00 in 1955 to 0.15 in 2008).

Patterns and Change in INCS
To illustrate the utility of the new measures, we provide three brief descriptive analyses to assess changes in density in the world polity (e.g., Boli and Thomas 1997) and in inequality across regions (e.g., Beckfield 2003) Figure 1 also records the density of the bipartite (country-to-INGO) network as a quantitative measure of connectedness.Calculated as the percentage of all possible connections in a network that are actually observed, density allows us to assess whether the growth in INGOs over time is actually increasing the number of connections between countries.Calculating density on the changing sample of countries limited to those with populations of greater than 1 million in 2000, we find that in 1950 the density of the network was 0.21, meaning that we observed 21 percent of all possible connections in the country-to-INGO network in 1950.Over the following 60 years, the density of the INGO network increased to 37 percent connectedness.
Assessing density on the changing sample could depress the estimate of density because we expect new countries entering the world system to have fewer ties to INGOs.Indeed, if we consider only the sample of countries sovereign in the international system since 1955, density increases from 0.29 to 0.47, so that ultimately 47 percent of possible ties are observed.Increasing connectedness is also evident if we consider the average number of INGO ties held by countries over time, which increases from 23 to 253.
Besides increasing density overall, world polity theory further suggests that we should see increasing equality in levels of state centrality, and especially less regionalization, over time.But Beckfield has contested this view in a series of articles looking at intergovernmental organizations (2003,2010).To investigate this claim for INGOs, Figure 2 presents variation over time in the INCS by region as well as the network density for each region.Regional network density assesses the level of connectedness among countries that are geographically proximate.
Clear differences are apparent across regions in average levels of INCS, variation in these scores, and trends over time.To begin, the West shows an increase in median INCS from 1950 (0.83) 19501955196519731978198319881993199820032008Africa, 1965density .17, 2008 19501955196519731978198319881993199820032008Asia, 1950density .20, 2008 19501955196519731978198319881993199820032008Eastern Europe, 1950density .34, 2008 19501955196519731978198319881993199820032008West, 1950density .55, 2008 19501955196519731978198319881993199820032008Latin America, 1950density .23, 2008 19501955196519731978198319881993199820032008Middle East, 1950density .17, 2008 density .33  scores in Africa.In sum, sub-Saharan Africa does not show evidence of overall increasing integration into the network.Asia, Latin America, and the Middle East and North Africa show some improvement over time, with continuing significant variation in INCS.In all three regions, the median INCS has increased, with some countries achieving very high INCS, but all three regions also show the bulk of the distribution of INCS remaining below 0.6, even in the latest years.
The same patterns are evident when looking at regional density-that is, the density of the bipartite (country-to-INGO) network within each region.The West is much more highly interconnected, with 79 percent of possible ties in 2008, compared to all other regions.In 2008, 35 percent of possible ties are observed in Asia and Latin America, and only 21 percent of possible ties are observed across sub-Saharan Africa.The 2008 pattern maintains inequality that was already established in 1950.
In 1950, the West had 55 percent of all possible ties, compared to Latin America with 23 percent.
Eastern Europe shows a unique pattern, largely because of country births and deaths.Splitting the results into Cold War and post-Cold War eases understanding of the Eastern European boxplots.From 1950 to 1988, Eastern Europe demonstrates a significant increase in INCS and increasing regional density (from 34 percent of all possible ties in 1950 to 55 percent in 1988).By 1998, however, it is clear that the approximately 20 additional countries entering the international system are generally less integrated into world society.Density scores decline to 28 percent of all possible ties.The USSR successor states have generally not reproduced prior levels of integration.
Figure 3 presents variation in the INCS over time for the constant sample of countries sovereign between 1978 and 2008.Patterns are similar to those for the changing sample, with the West coalescing around high INCS, sub-Saharan Africa remaining largely unintegrated, and Asia, Latin America, and the Middle East and North Africa showing small gains.The Eastern European countries that were sovereign throughout the period (i.e., excluding the USSR successor states) show a very different pattern than in Figure 1.These countries significantly increase their median INCS over time, even though variation remains pretty consistent over the period. 8 The density of the bipartite (country-to-INGO) network within each region is also fairly similar when we consider only countries that have been sovereign continually since 1978.The West is again much more highly interconnected, with 69 percent of possible ties in 2008, compared to the 36 percent of possible ties observed in Asia, 38 percent in Latin America, and 22 percent in sub-Saharan Africa.Without the newly sovereign states to bring down overall levels of density, however, Eastern Europe continues its upward trend of increasing density, reaching 62 percent of all possible connections by 2008.
In short, looking across regions we find that regional disparities are consistent across time.The West always leads, followed by Latin America, Asia, the Middle East and North Africa, and finally by sub-Saharan Africa.In general, although we observe gains in network scores and density within regions across time, regional stratification persists.Finally, we can consider individual country trajectories.Figure 4 tracks the INCS versus raw INGO counts of Chad, Thailand, the United Kingdom, and Spain over time.Beginning with consideration of the trends in the INCS, Chad enters the international system after independence in 1960 with very low connectedness to the world polity.Over the subsequent fifty years, its connectedness barely increases, reaching only 0.15.In contrast, not only does Thailand enter our dataset in 1950 with a higher level of connectedness than Chad, but its embeddedness in the world polity increased dramatically from 1950 to 2008.Even so, Thailand has yet to approach the levels of connectedness displayed by the United Kingdom in all observed years of our data.Spain, the fourth country in the figure, is an interesting case as a Western country with very low levels of connectedness in the earliest periods but substantial growth thereafter.The observed trend in INCS corresponds to the history of Spain in the international system.After World War II, under General Franco, Spain was both politically and economically isolated, but Cold War politics made Spain more strategically important to the United States in the mid-1950s.In December of 1955, Spain joined the United Nations.The jump in embeddedness corresponding to Spain's U.N. membership is remarkable.After Franco's death in 1975 and the restoration of democracy, connections increase further, resulting in Spanish levels of embeddedness matching the rest of the West.
Comparing the INCS to raw counts of INGOs shows significant differences in the trends over time.The raw counts of INGOs tend to increase for almost all countries as the number of INGOs in the world increases.In some cases, such as Chad, the conclusion about the place of the country in the world polity is similar.In others, a distorted picture results from looking at the INGO counts.For example, the INGO counts suggest exponentially increasing connectedness for the United Kingdom over time, whereas the INCS reveal that the United Kingdom was among the most connected countries in the world polity from the beginning of the period.Similarly, Spain's rise in embeddedness, with diminishing returns after 1990, is masked in the raw counts.Overall, exponentially increasing numbers of INGOs make it appear that almost every country in the world is increasingly connected to the world polity when only counts are considered.When the entire network is considered, it becomes apparent that some countries have always been deeply embedded in the world polity, and others have remained outside the core of the world polity and do not increase their connectedness over time.Hughes et al. (2009) argued that capturing the emergent properties of the world polity requires network analysis, rather than a focus on INGO counts.World polity theory is a relational theory; it implies that ties between states through INGOs create an international social network.Rather than states' behaviors being driven solely by internal attributes such as their GDP, it is the relative positions of power among states that shape common beliefs and influence state behavior. 9Tests of world polity theories that consider INGOs as membership counts alone have thus not gone far enough in capturing how asymmetries and inequalities in the overall network are likely to influence states' behavior.In this article we extended the INCS in a number of ways.We significantly extended the time period considered, from a thirty-year period to a sixty-year period .We also created four relational measures of connectedness to the world polity that use different samples of countries.The first measure uses a worldwide sample of countries at each time point.The measure addresses changes in the world as countries break apart to form new nations and thus incorporates the dynamics of a changing world system.A second measure, reported in Table 1, is restricted to countries with populations greater than 1 million in 2000 but continues to allow for country births and deaths in a changing sample.For other research questions, scholars may be especially concerned about achieving comparability across certain countries across time.Because countries typically enter the world system with fewer INGO memberships than established countries, incorporating new countries alters the scale by changing which countries are the least connected.

Conclusion
Thus we also provide the INCS for two constant samples of countries, one with countries that were consistently in the international system since 1955 and one with countries consistently in the international system since 1978.We provide INCS for all samples in the appendix.
To demonstrate the utility of the expanded measures, we provide a visual representation of country connections through INGOs mapped onto world geography.The visual picture of the entire network of INGO ties provides a general overview of the structure of the world polity network and shows that it is highly dense.But these network graphs also demonstrate inequality-some countries are more embedded at the core of the network, and others remain peripheral.We also confirmed regional inequality in the world polity at all time points.Sub-Saharan African countries have the lowest average INCS, with little growth over time.In contrast, Western countries have consolidated their initial highly central positions.The regional disparities we observe are consistent over time, a finding that contradicts earlier views that the world is becoming more equal (Beckfield 2003).
Previous work has used the exponentially increasing rate of INGO foundings to argue that the world is increasingly interconnected (e.g., Boli andThomas 1997, 1999).Using our network-based measure, we find that the world is not uniformly increasing in connectedness.Instead, even though INGO foundings continue to increase and the number of raw ties between countries is increasing, regional differences in density persist, and some countries remain highly disconnected.Thus we demonstrated that exponentially increasing possible ties between countries through INGOs does not translate to actual ties between countries over time.This trend is likely to continue, as newer organizations are no broader in reach than established INGOs.
Implications of the extended and expanded INCS are myriad.States that are more central in the network likely have greater authority over not only the diffusion but the creation of policy scripts.Powerful, more connected states are more likely to influence the norms and ideologies that INGOs use to create and disseminate their agendas than weaker, more peripheral states.States with more power may also have greater freedom to ignore policy scripts that do not mesh with domestic agendas.Thus the structural position that a state holds in the overall network will influence policy choices, including the choice to ratify human rights conventions, enact environmental laws, approve quota or any other outcome commonly studied by world polity scholars.Just as Ingram, Robinson, and Busch (2005) demonstrated that using an intergovernmental organization network improves our understanding of bilateral trade, or Hafner-Burton and Montegomery (2006) found that intergovernmental organization ties between states reduced conflict, it is possible-indeed likely-that using network-based measures of INGO connections may alter our substantive findings on the dissemination and adoption of policy scripts. 2 USSR/Russia is combined as an exception, due to its geopolitical importance.
3 If they were in the first year (1950) or last year (2008) of the data, we allowed imputation if the INGO had at least one adjacent year of data (1955 or 2003, respectively).For all other years, INGOs were imputed if they had at least one preceding and one year of data, which need not be adjacent.
4 For the last year of the data, 2008, we required two adjacent immediately prior years of membership data (1998 and 2003).We did not impute INGOs for 1950, since we cannot realistically project membership backwards for organizations that did not appear in the yearbook.
5 We also imputed, again using a mixed-effects Poisson model and Bayesian prediction, INGOs for countries in their first year of sovereignty.Boli and Thomas (1999)

Figure 1 :
Figure 1: Country to Country Ties, 1950 and 2008 to 2008 (0.93).7 The variation in INCS in the West sharply decreases, with the first quartile increasing from 0.72 in 1950 to 0.88 in 2008.Whereas France held the top position in the INCS distribution from the beginning of the period, other countries in the West are increasingly integrated into the global network.Overall, the pattern in the West is one of strong consolidation around high INCS.In contrast, consider sub-Saharan Africa.Although a few countries have achieved high INCS, the bulk of the distribution remains below 0.3, even in 2008.Indeed, the minimum observed INCS in the West are similar to the maximum observed sociological science | www.sociologicalscience.com

Figure 2 .
Figure 2. INCS Scores by Region over Time, for countries with > 1,000,000 population

Figure 2 :
Figure 2: INCS Scores by Region over Time, for Countries with > 1,000,000 Population Notes: Results use samples limited to countries with populations of greater than 1,000,000.Boxplots for Africa in 1950 and 1955 are not presented because only three sovereign nations existed in Africa in those years.

Figure 3 .
Figure 3. INCS Scores by Region over Time, for constant 1978-2008 sample

Figure 3 :
Figure 3: INCS Scores by Region over Time, for Constant 1978-2008 Sample

Notes1
For early time points, we relied on organization names, aims, structure, and country memberships to select organizations.Beginning in 1983, we made use of UIA codes sociological science | www.sociologicalscience.com to select both B (universal) and C (intercontinental) organizations.The set of INGOs with potential universal membership is smaller than the full set of INGOs that includes regional INGOs and is typically used in count measures.

Table 1 :
INCS . We begin by comparing a map of the network structure (overlaid on geography) in 1950 and 2008.Darker lines indicate increasing numbers of connections between states.Growth in INGO ties over time is immediately apparent-the network was less dense in 1950 than in 2008.But the differences in embeddedness across countries and regions that appear in 1950 are strengthened in 2008.Europe is the densest region and most central in the overall network in both years.
Hughes et al.'s INCS ranks countries by document lags in the UIA data that suggest that countries, in the years immediately following sovereignty, have lower than expected INGO counts.Detailed information on our imputation strategy is available from the authors.6 Two related international network measures exist for intergovernmental organizations (IGOs) and trade.IGOs are different from INGOs because they are formed by treaty and have membership composed of states, rather than individuals, groups, or businesses (Pevehouse, Nordstrom, and Warnke 2004:103).Hafner-Burton and Montegomery (2006) calculated country centrality in the IGO network through 1992.In 1988, the correlation between the INCS measure and IGO centrality is 0.73.De Benedictis et al. (2013) calculate eigenvector centrality scores based on international trade flow for 1955-2010.In 2008, the correlation between the INCS measure and trade eigenvector centrality is 0.77.