Notes on: Young, K., Goldman, S., and O'Connor, B. (2020). How White is the Global Elite? An Analysis of Race, Gender and Network Structure.  Global Networks. 124.

Dave Harris

They intend to use network analysis to look into relationships among elites and they are especially interested in diversity of board members of 'large corporations, think tanks, international organisations, and transnational policy planning groups'. They are also interested in the '"core – periphery"' hypothesis, which predicts that nonwhites and women will achieve elite positions but will then be marginalised to the periphery, leaving the core as 'significantly more white and male'. White men actually increase in representation in the core.

It is obviously important to look at the composition of elites in terms of decision-making and the setting of policies influence the global economy. Members of boards forged ties across organisations often through governing personnel themselves. Corporations are connected also to areas of governance which means they can inform political strategies.

This article pursues an empirical examination board of members of key organisations and their interrelationship, the social networks as well as simple head counts. They also evaluate the core periphery hypothesis. They do this through research methods involving constructing network measures and other descriptive statistics.

They argue that the diversity of global elites matter because 'the background of individuals has the potential to inform their governing behaviours' (3),[they cite a number of other studies suggesting that the composition of leaders action does matter]. Secondly diversity affects organisation performance, so that 'greater gender diversity… Is associated with positive effects on organisational performance… Racial and gender diversity within corporate management appears to enhance the financial performance of firms'. Thirdly there are issues of representation and legitimacy, contentious political issues, especially for corporations with global reach, and organisations developed to insist on EDI. Fourth the composition of global elites helps understand globalisation and the distribution of wealth, especially if networks produce variable and unequal outcomes.

Factors affecting in-group solidarity can also include being Protestant (in the USA) (4). There can be changes over time as various groups struggle to be included, and some marginalised groups have indeed been included, including women, blacks, and Jews.

The North Atlantic has been the centre of global networks until recently, favouring English-speaking. The rise of Japan in the 70s and 80s meant that Japanese corporate elites had to be included into North Atlantic networks, and there is now a shift of equilibria to the east, to China, Southeast Asia and India, producing a possible 'separate "East Asian" cluster', with differential integration of elites into global networks.

There are demographic shifts as well, including some within national societies, including racial and gender diversity and counter reactions to it. The actual state of diversity is still relatively unknown.

It is still difficult to study the dynamics of elites especially forms of inclusion and exclusion. Changes of diversity are slightly easier, including the inclusion of particular minority groups such as Jews, blacks and women [lots of references], but they measure diversity in terms of total numbers, and usually only within American corporations. The study looks at a broader terrain.

Race and gender specifically have been analysed in 'a few exemplary studies' [cited on 5], but only one has focused on diversity within the global elite — a study of attendance at conferences of the Trilateral Commission over time. This is still only one of 10 transnational policy planning groups. It did find that 'women and nonwhites are generally excluded from the inner circle', defined as those with long attendance records.

They use network analysis and focus on the elite corps or inner circle, that 'group of individuals who not only have superordinate control over resources at their command… But who also occupy an especially prominent place within a neat social networks'. This core helps maintain the class identity with the elite, because it relates to the management of particular capitals — so the interests of Walmart is 'moderated by the interest of capital in general'. The same goes for corporations political roles, which takes the form of 'moderate and pragmatic political views… [Which protect]… The status quo'. There is a higher degree of coordination among the inner circle, shown by studies which include 'elite sociality'. We can also make core- periphery distinctions and many studies have [lots of references 6]. The core is useful in establishing 'social exclusion and insider-ism that pervades social life'.

We can assess how 'proportionately representative' different race and gender groups are. White males can expected to be the most well represented with the strongest mechanisms of in group solidarity, reflecting their historic dominance. It's easy to think that white males social networks marginalised blacks or East Asian females. There may be incentives for tokenism, however or 'other strategic considerations', or even genuine changes in norms. The dynamics may be complex.

However it is unlikely that the inner circle change, and they have been shown to be resilient in earlier studies. Corporate organisations and their networks do very across countries [loss of references to Scott 1991 here] and there are specific studies on Chinese Latin American or Anzac elites. Studies of the North Atlantic found some similar properties — a study in the notes compared network properties to random networks in order to measure 'how "clubby"' elite networks were. They found no substantial difference with ordinary board members, but there was more difference among those corporate board members on multiple boards.

They started their study by collecting globally powerful organisations as above. They chose a top list of global corporations by looking at total revenue rather than assets, in order to avoid overemphasis on financial corporations. They included 30 corporations. They chose think tanks according to their assessment of the role played — they ended with the 'most prestigious 30 think tanks in the world' using some 'widely used' league table produced by the Lauder Institute at the University of Pennsylvania. They looked at transnational planning groups like the World Economic Forum and the Trilateral Mission which have an important role and included the boards of the 10 most prominent ones, drawing on 'existing scholarship'. They added the governing boards of six major international organisations, things like the Bilderberg Conference, and various Round Tables, and various NGOs, especially the top 10 globally ranked 2015. They collected the boards of all these organisations through annual reports and then searched websites and others for their personal connections. Overall they had 96 organisations, possibly over representing certain regions of the world, but nevertheless those with the global remit. They are aware that they chose organisational leaders rather than other members of elites such as extreme wealth holders who do not appear on governing boards. Overall it's a larger sample than in previous elite stones.

They ended with 1600 key individuals and explored the networks among them as a base network. They have probably underrepresented the number of ties. There were some quality checks, checking on ambiguity and doublechecking. They admit degrees of error nonetheless. They identified modes and individuals, and 'edges' which connect individuals to organisations, producing both a base organisational network and an extended organisation network. They analysed the network in terms of statistics such as density, average path length, modularity and transitivity. [Table on 10 — generally the extended organisational network had far more nodes and edges, but slightly lower path lengths modularities and transitivities — but much lower density]

They coded the race of members using 'standardised headshot pictures', recognising that perceptions can vary and that these categories are subjective. Other work use different codes, some referring to local context. They used 'the most generalisable categories that could be reliably coded' they did intercoder reliability checks on their classifications and found reliability to be high [the categories were black, East Asian, Hispanic Middle Eastern, South Asian, and white — white men were 54.5 of the elite, East Asian women 1.64%, Black men 4.15%, Black women 2.45%.Lowest of all -- Hispanic women 0.44%.. They estimated gender through head shots as well, with known risks.

They then conducted a 'k– coring exercise' A k-core. 'Is composed of all of the notes that continue to be included in a network, when only those with at least K connections are included' (12). So if we are interested in a '4 – core' network, we eliminate all nodes with three or fewer connections. [The point is we can then see set of individuals on the boards of four or more organisations]. They do the analysis for different values of K and found that the maximum K score was 2 for the base organisational network, 43 highly interconnected individuals, and for the extended organisational network, the value was 4, with 151 highly interconnected individuals. [The graph shows gender differences as well, with understandably more male representatives than female, except for Black, Middle Eastern and South Asian groups, where they are quite close. White members have quite different results].

White males are clearly the dominant group and their representation increases as the K score increases, in other words as we move from periphery to core. White women are the second highest represented group, but their representation declines as the K score increases, indicating that it is a combination of race and gender that excludes them relatively.

Nonwhite categories, all individuals, whether male or female decrease as the K score increases with the base organisational network, and for the extended organisational network this is still true, despite some marginal increases from 3 to 44 black men and East Asian men this is 'a double-edged sword' — with small numbers of individuals we might find increasing diversity, but small numbers also mean 'lower confidence in observed variance' (13).

We can develop more tests, allowing for the fact that we have different sizes of board — [otherwise, apparently there is a risk of 'seeing individuals on large boards as more highly connected than individuals on small boards regardless of their degree of collectivity across organisations'] (14). Instead they use a weighting method, increasing the weighting of size across organisations compared to ties within an organisation — to individuals that common size across all our organisations of a weight of four, whereas ties within one organisation are given a weight of one. This helps recalculate minimal distance parts, since the latter have less path to travel [and there's something about edge weights which I don't follow]

Absolutely superb network diagram, a colourful cluster ensues on page 15, using new methods to examine the core periphery structure of the network. They are lovely! These network analyses have been used in studies other than elite networks, like global trade networks or US airport straight analysis again it seems a matter of using information from 'weighted edges to ascertain the backbone of the network'. Yet another method has been used to study the Danish elite and 'conforms extremely well to notions of an "inner circle"' on the basis of multiple ties and also structural properties of the network, minimal distance parts resulting in a score given to individuals who can reach other individuals more efficiently — reachability here involves 'inverted edge weights'.

They are happy that these beautiful diagrams show that 'white men are the only group with substantially higher representation in the core than the periphery' (16 – 17). There are exceptions — black women either hold the same level of representation in the core and periphery or else increase their representation in the core' (17) although this is not consistent across the different methods. Also 'the relative change for white women is very small' using one method, but 'more evident' using other methods.

They investigated the consistency of results across the methods by trying them out on their base organisational network, amended and reduced to those with only two connections. This would assess the robustness of their findings. They used a different way to depict change in the core to the periphery, 'averaged across the three core periphery methods' (18) used in the above section. There is more variability here with directional consistency for some groups, but not all. East Asian women and white women are less well represented in the core. White men are increased in representation, in a consistent result. Inconsistencies might be produced by smaller numbers of individuals of the women.

There is still a need for further research, especially on 'racial homophily and the operation of gender norms in professional life'  (20). There are other elite attributes such as 'age, elite education and skin tone' which mark status. There may be changes underway, including in the geography of wealth. There is need to investigate the causal mechanism underlying these dynamics.