Who do sovereign wealth funds say they are? Using structural topic modeling to delineate variegated capitalism in their official reports

Abstract

Sovereign Wealth Funds (SWFs) are an emerging community of investment organizations, subject to the conflicting demands of international and domestic institutions. From the lens of variegated capitalism, one can explain their diversity and apparent contradictions by underlining how they embody negotiations between global and local institutional pressures. Many SWFs now publish official reports, which represents a unique opportunity to analyze how they portray these institutional influences. We use an unsupervised machine learning method, structural topic modeling, to analyze the 2015 official reports of 40 SWFs and 17 public pension funds globally. With this novel method, we propose the first international, community-wide study that shows how capitalism is adapted across regions, political systems, and even organizational frontiers through SWFs. We show that region and political systems are factors supporting a polarization of missions between neoliberal and state capitalist view. However, this is nuanced by international clubs that support new investment strategies, and some high-profile SWFs that mix the neoliberal and state capitalist views.

Keywords

Variegated capitalism sovereign wealth funds machine learning structural topic modeling professionalization long-term investors

Introduction

“Sovereign wealth funds” (SWFs) can be defined as “government-owned investment funds operating in private financial markets.” (Monk, 2009: 1) This definition underlines their unique position at the intersection of “high finance and high politics”, as they use international financial markets to grow national wealth (Backer, 2009; Drezner, 2008). However, SWFs are also a very diverse group of organizations, with assets going from a few millions to billions of dollars, and missions that cover macro-economic stabilization, pension reserve, national development, and maximization of capital coming from natural resources or industrial exports (Clark et al., 2013; PWC, 2016). Their diversity in size, origin of capital, missions and investment strategies can be explained through the lens of variegated capitalism (Peck and Theodore; 2007). When confronted with social constraints nationally, the ideal of neoliberal finance and free markets is adapted, leading to a mosaic of capitalisms across the globe representing these various national adaptations (Haberly, 2011; Haberly and Wójcik, 2017).

However, despite this diversity, SWFs have been organizing as an emerging professional community. Indeed, SWFs have been increasingly co-investing together and with other institutional investors, as well as creating and joining international professional organizations (Bachher and Monk, 2013). If the first one, the International Forum of SWFs, was strongly influenced by an existing neoliberal international organization, the IMF, recent ones (such as the Long-Term Investors Club) were created and ran by SWFs themselves and invited other investors, most notably public pension funds (PPFs), around the ideal of long-term investing (Monk et al., 2017). These platforms and growing engagements with each other lead us to think that groups of SWFs sharing similar views on their missions and how they should invest are emerging. SWFs could therefore now be best understood as building up their identity at the intersection of global finance, national politics, and the emerging professional identity of the long-term investor.

So far, exploring the identity of SWFs in terms of missions and visions needed to rely on interviews and has mostly been achieved through single or small-n case studies, making the task of comparing SWFs very difficult (Backer, 2009; Clark et al., 2013; Clark and Monk, 2012a). However, SWFs have increasingly communicated through websites and annual reports on their mission, vision and investment strategy, even if the publication of financial results is scarcer. The PPFs with whom they co-invest, and whom they join in international organizations, also publish annual reports. This has created a corpus of texts that can be used to compare SWFs’ declared identity, missions and values. In addition to this availability of data, unsupervised machine learning techniques can now assist qualitative researchers in finding common topics across a large amount of texts in less time. The structural topic modeling method allows to find these topics and to measure the correlation between the prevalence of a topic in a text and external variables.

This paper uses this technique to identify topics used by SWFs and fellow long-term investors to describe their missions and investment strategies. In addition, it shows which external factors, notably belonging to a professional association or a geographic region, is correlated with the prevalence of topic. This paper presents the first large-scale taxonomy of SWFs based on their declared identity, while whereas existing typologies were based on a lower number of case studies or on surveys. We offer a map of SWFs’ and PPFs’ goals and investment strategies. We also, show how these echo values from global neoliberal finance, state capitalism and the emerging professional identity of the long-term investor. Finally, measure which institutional factor influences where SWFs sit on this map.

We first describe the institutional forces influencing SWFs using the lenses of variegated capitalism and institutional emergence. These ground our hypotheses about the topics we expect to find in SWFs’ official reports and the external factors likely to influence them. We then describe the new method that is structural topic modeling (STM). The fourth part presents the results of the STM analysis. We conclude with the implications of our results on our understanding of SWFs’ identity, and what doors STM opens for Economic Geography.

Review of the literature

SWFs between global and national institutional influences

Sovereign Wealth Funds are at the intersection of two worlds. As investors on global financial markets, they need to comply with the rules of “global financial capitalism”, which comes from neoliberal financial markets originally created in the UK and USA, and whose rules and values are carried by international organizations such as the International Monetary Fund (IMF) (Monk, 2011). As safeguards for their national economies or retirees’ pensions, they also follow parallel logics, in line with their political and historic realities, and put the search for financial returns at the service of national development, economic stability or long-term value creation (also called capital maximization) (Dixon and Monk, 2011a). As such, SWFs have been described as an expression of resistance to, but also engagement with global financial capitalism, and a tool for nation-states to retain sovereignty against globalization forces (Dixon and Monk, 2011; Monk, 2011). In line with this thesis, several regional case studies have shown how SWFs in several locations bend some of the rules of neoliberal finance to protect social and political interests while still wanting to use markets to grow national wealth (Clark et al., 2013; Clark and Monk, 2012a). Studying SWFs as a community answers Peck and Theodore's (2007) call for studies of “variegated capitalism”, which go beyond the dichotomy between coordinated and liberal market economies proposed by the Varieties of Capitalism theory (Hall and Soskice, 2001), and account for the complex co-evolutions between national and global forces. This lens describes capitalism as a polymorphic process, with uneven spatial and temporal development, where the rise in homogeneity of systems of production and exchange does not erase territorial specificity and institutional diversity (Dixon, 2014; Peck and Theodore, 2007). The same international pressure from neoliberal ideals will therefore produce diverse outcomes across the globe. But documenting capitalist development paths at a global scale and going beyond national-level case studies should lead to recognize patterned relationships between local capitalisms and could help identify factors beyond the national level to explain local adaptations (Haberly and Wójcik, 2017; Peck and Theodore, 2007).

The process of transformation of global norms when confronted with pre-existing national institutions is apparent in Mehrpouya’s study (2015) of how SWFs dealt with pressures to follow the rules of global financial capitalism during the negotiations of the Santiago Principles under the lead of the IMF in 2009. These principles ask for transparency into SWFs governance and investment policies in line with the rules of neo-liberal capitalism. The transcripts of the negotiations reveal several blocks of SWFs pushing different visions inspired by their own national political and cultural situations. Mehrpouya (2015) underlines that “transparency” was contentious because of its different local implications for SWFs. Norway’s SWF saw it as a public right and necessary component for SWFs. For Singapore, it was a market instrument necessary for international legitimacy but threatening long-term investment. Finally, for Middle Eastern SWFs, transparency could open claims for capital from domestic actors and destabilize their regional geopolitical environment, justifying their need to transform this requirement (Mehrpouya, 2015). This description supports the view that SWFs will bend the global ideal of capitalism to their local political and cultural setting. This means SWFs could be grouped in several groups sharing common values according to geographic or political lines.

Refurbished state capitalism as a new global institutional influence

SWFs also show patterns of resistance to the standard rules of global financial capitalism as a global community. This goes hand in hand with Haberly and Wójcik's (2017) observation of a dramatic relegitimization of state ownership, but a state ownership that seeks to achieve a balanced synthesis of the defense of national sovereignty and global economic integration. In this “refurbished state capitalism”, market forces are judged to be good as long as the State can control some key economic levers (McNally, 2013), and this vision seems to gain more legitimacy across regions (Haberly, 2011; Haberly and Wójcik, 2017).

This contradicts the thesis that SWFs adopt international financial rules only as a façade to gain legitimacy but do not conform with them (Balding, 2009). Instead, it proposes that SWFs create their own mix according to their own value system of a state-controlled capitalism. It goes against a common argument that SWFs are “decoupling” their words from their behaviors (Scott, 2008), which happens when the values carried by powerful institutions are impossible to put into practice locally, but that recognizing it publicly would threaten the legitimacy of an organization and endangers its activities and survival (Dixon and Monk, 2012; Truman, 2010). SWFs are creating a middle way, adopting honestly some principles from neo-liberal finance but also leaving others aside. They routinely adopt performance incentives for top managers, use advisory services by international fund managers, and restructure to improve their corporate governance (Dixon and Monk, 2011b). These new state-owned entities invest on international markets and follow their rules and do not obstruct the operation of market forces nationally, thus being “one of the most globalization and liberalization-friendly interventionist instruments” (Haberly and Wójcik, 2017: 252). In addition, this openly admitted alliance of national and financial goals is not restricted to developing countries and commodity exporters but has been adopted by large developed countries with a history of state intervention such as France and Italy.

SWFS’ professionalization around the values of long-term investing

SWFs can be a vehicle for new global institutional values to emerge and spread across countries. Indeed, Djelic and Quack (2003a) underline that influence between institutional levels can be top-down (‘trickle-down’), coming from the transnational level to a lower level (nation or organizational population), or bottom-up (‘trickle-up’), coming from nations to the transnational level. They propose that actors from lower levels, and notably organizations, can combine institutional fragments at the transnational level using three different modes, dominant, negotiation and emergent.

In the dominant mode, the building of institutions at the transnational level reflects one dominant local or national model. This is the case for economic rules such as shareholder values, strongly influenced by the Anglo-Saxon model (Kleiner, 2003; Tainio et al., 2003). In the negotiation mode, institutions are the result of confrontation, debate and bargaining between actors coming from different national rule systems. The results of the negotiations reflect the relative powers, notably through access to resources, of actors involved in the negotiation. This is the case in international organizations such as the International Monetary Fund. Finally, the emergent mode is driven by actors that are not nation-states but individuals or organizations, and therefore exhibit more transnational identities. This mode takes place at the borders of multiple rule systems, and the outcome is harder to predict. It is also a way for organizations to create institutional change, modifying the rules, values and habits of their industry or community (Fligstein and McAdam, 2012). While the first two modes were dominant during the 19^th and first half of the 20^th century, there seems to be an increasing use of the emergent mode since the 1950’s through professional organizations (Djelic and Quack, 2003a, 2007). Professional associations have been shown to be powerful instruments of institutional change, helping organizations deflect international and national institutional pressures and giving way to new value systems such as corporate social responsibility (Greenwood et al., 2002; Meyer et al., 2013).

SWFs increasingly participate in international clubs and associations that gather multiple types of institutional investors. For example, the Long-Term Investors Club (LTIC) was created by the French and Italian SWFs in 2009, and still gathers SWFs, development banks, and pension funds at their conferences. The Institutional Investors Roundtable, “a platform for and by long-term investors”, was created in 2010 by a group of 16 leading pension and sovereign funds, among which are the SWFs Temasek (Singapore), Abu Dhabi Investment Authority, the Russian Direct Investment Fund, as well as the Canadian pension funds BCIMC and CDPQ, and the Dutch pension fund PGGM. These “clubs” gather institutional investors with diverse missions, from authoritarian and democratic regimes around the world and could therefore be a vehicle for SWFs to expand the boundaries of their organizational category. So far, they seem to have helped SWFs and PPFs to take part in joint co-investment platforms such as OMERS- sponsored Global Strategic Investment Alliance, and TIAA-CREF-sponsored Agriculture LLC (with Swedish SWF AP fund, Canadian PPFs CDPQ and BCIMC) (Monk et al., 2017).

These clubs could also help SWFs claim and legitimize a new set of values and investment practices, different from those of the Santiago Principles, and less linked to a specific world region or political regime. Through this emergent mode of institutionalization, SWFs could push back against both national political influences, and global capitalist dogma. SWFs rally professionally increasingly under the name of “long-term investors”, rather than SWFs, benefiting from increasing calls for long-term investing from the OECD, which defines it as “patient”, “productive”, and “engaged” profitable capital (G20/Oecd, 2013).

This new identity comes with the values of supporting greater financial stability through less pro-cyclical investment strategies and longer-horizon holding periods; invest in projects and companies essential to economic development such as infrastructure, SMEs and green growth initiatives; and be active shareholders enforcing better corporate governance in the companies and projects in which they invest (Papaioannou et al., 2013). Long-term investing has also been defined based on the investment strategy available to investors whose liabilities and return objectives run over decades. These definitions include: the absence of non-financial reasons to sell or buy assets and therefore the possibility to lessen the impacts of financial crises on the investor’s capital, also called “discretion over trading” (Denison, 2010; Warren, 2014)); and the focus on “value instead of price” (Kay, 2012; Warren, 2014), targeting cash flows over the business cycle and paying less attention to interim asset prices. Some principles of long-term investing are close to state capitalism, notably in their preference for real assets that contribute to the sustainable development of countries (Gelb et al., 2014). Long-term investment is also often associated with direct investments by institutional investors, alone or through co-investments, but without traditional financial intermediaries whose rewards have traditionally been linked to short-term results (Clark and Monk, 2012b).

A structure in the community of SWFs at the intersection of multiple institutional influences

SWFs’ missions and investment strategies are influenced by a complex set of institutional factors, including dominant modes of influence from neoliberal institutions regulating markets, nations’ equally dominant influence as owners of SWFs’ capital and wishing to use it to protect their national economies, but also emergent influences among a larger set of institutional investors interested in becoming more profitable and sustaining financial crisis thanks to long-term investing principles. Looking at a large set of SWFs across the globe and their close collaborators in international clubs can help us get a picture of the structure of this community, likely to portray all these influences.

In line with the three modes of institutionalization we described, we expect that our set of SWFs will include:

Organizations strongly aligned with neoliberal finance when their national values and rules align with this global system;

Organizations supporting the values of refurbished state capitalism, notably when their political regime is authoritarian and has a history of state capitalism;

Organizations that have adopted the long-term investment stance, most likely among organizations that take a leading role in international professional associations of long-term investors.

The use of a novel algorithm-based methodology can help us see the structure of the community of SWFs in a way that challenges and informs our theoretical notions. We describe this method in the next section.

Using structural topic modeling to analyze the content of official reports

This study aims to answer the question: “how do SWFs described their missions and investment strategies in official reports in 2015–2016, and what institutional factors were associated with different descriptions?

To answer these questions, we make three important choices that we discuss below: we analyze a population of long-term investors that includes SWFs and some PPFs; we use official statements about missions and investment strategies in annual reports and websites; and we use a new machine learning method: structural topic modeling, that we think offers great promise for economic geography.

Analyzing SWFs and some PPFs

There isn’t one list of SWFs that one can use to study the whole community. The two most complete lists we found were that of PwC’s “Sovereign Investors” (2016) and of the Sovereign Wealth Center. We completed these lists with members of four international organizations gathering many SWFs (Monk et al., 2017) because we were interested in the influence of professional groups on SWFs. We also added a list of sovereign development funds identified by Clark and Monk (2015) to make sure this variation of SWFs was included. From this list of 167 organizations from eight sources, we supposed some organizations would be misclassified, and retained only the 83 organizations listed by at least two sources (Appendix 1 lists these organizations), which include 66 sovereign wealth funds and 17 public pension funds. Out of those, 57 organizations published information akin to an annual report (40 sovereign wealth funds and 17 public pension funds). We kept PPFs in the list because the frontier is sometimes thin between a PPF and a SWF with a mission of pension reserve, and to be able to measure the influence of professional associations on both SWFs and PPFs that were part of these organizations.

The value of official reporting

SWFs’ description of their missions and investment strategies in publicly available written documents gives us access to their organizational identity, as defined by the characteristics an organization gives to itself (Albert and Whetten, 1985). These public statements are by essence strategic and political, as they are written for a purpose and an audience. In our case, they can be a way for SWFs to gain legitimacy toward an international audience (complying with the international standards of transparency and accountability), to a national audience, or a professional group. In official reports, SWFs can use values and categories that have emerged in the community, ensuring legitimacy to a broad audience, but they can also support new categories that can create differentiation and a strategic position against competitors and alongside partners (Khaire et al., 2010).

We chose to study annual reports because they have a set template but also offer space for organizations to differentiate themselves, notably through their mission statements (Leuthesser and Kohli, 1997). In this study, we don’t discuss how close to reality these statements and reports are, or if they have a goal-setting value. Instead, we consider annual reports as public representations of organizations that can give them legitimacy and/or competitive agency, and position themselves among a set of potential partners, competitors and regulators (Argenti and Druckenmiller, 2004; Esrock and Leichty, 1998).

To position organizations among their peers, Albert and Whetten (1985) proposed to list all the characteristics of several institutionalized types of organizations, and position organizations on this spectrum. Adapting this idea to the realm of the values expressed and vocabulary used by organizations to describe themselves, we propose to use a method of computer-assisted text analysis to assess SWFs’ identities among their peers.

Structural topic modeling

Computer-assisted text analyses methods greatly reduce the cost and time needed to analyze large collections of text and are increasingly used in social sciences (Grimmer and Stewart, 2013). Beyond their efficiency, unsupervised algorithms can also help the researcher discover new themes. They are opposed to supervised learning techniques, which try to reproduce on new datasets the classification schemes learned on previously coded data. On the contrary, unsupervised learning methods can cluster data based on structural similarities with minimal input from the researcher. The algorithm does not assign a value or meaning to these clusters, thus letting the researcher assess the relevance and meaning of the newly found associations. While this method does not guarantee to systematically yield meaningful findings, it forces the researcher to take a new look at the data at hand and can help identify themes or influential characteristics that were not previously described in the literature.

Because we wanted to inductively find the themes used by SWFs and which organizations were closer to which, we chose to use an unsupervised algorithm called “structural topic modeling” (STM), which builds off the method of Latent Dirichlet Allocation (LDA). LDA is built on the intuition that each document will be a mixture of a limited number of topics, with a different distribution of these topics. The researcher sets the number of topics he/she thinks there are in a collection of texts and runs the algorithm on them. The LDA method creates topics by showing the words that are frequently used together in the texts (Grimmer and Stewart, 2013). This is based on the idea that humans infer that a text is about a topic by identifying the use in one paragraph of several words that pertain to the same lexical field. The LDA algorithm identifies those words that are likely to belong to a lexical field or topic based on their frequency of appearance together, and lets the researcher decide if they do form a lexical field (or not) and name it.

The algorithm gives as many sets of words with their frequency and co-occurrence as number of topics given by the researcher. The algorithm also gives a measure of the prevalence of each topic per document and across the corpus of documents.

The researcher has to interpret what each group of words is about based on these word frequencies, the words themselves, and by reading the documents that predominantly feature one topic. Blei et al. (2010) underline that it is an alternative to manual coding in which a human reader assigns a topic to a sample of text and get a sense of the meaning of the document by looking at its words and lexical fields to identify topics and draw parallels between texts.

In this study, we use the STM algorithm available through the ‘stm package’ in R (Roberts et al., 2015a). It expands on LDA and Topic Modeling by allowing researchers to add external descriptors to each document, and computes correlations between these external characteristics (covariates) and the prevalence of topics in a document or the word rate in each topic (structuraltopicmodel.com, Lucas et al., 2012, 2015). Here, we are interested in the weight of external characteristics such as regions, political systems, size of assets under management, on the proportion of topics in each text to refine the results. We use these results to group SWFs that use predominantly some topics in their texts, although this clustering of documents is not a direct result of the method. This is important as it differentiates this method from clustering methods that create groups of points based on covariates. Here, the method is really focused on creating topics based on the frequency with which words appear together in all texts. Then it measures the prevalence of topics within each document and the correlation between this prevalence and covariates. The analyst can include a covariate when he/she thinks it can influence how much a text is likely to discuss a topic. The algorithm does not force all topics according to these covariates though, and it includes “shrinkage priors” or regularization, that limit the risk of overfitting to a covariate (Roberts et al., 2015b). Ultimately, the covariates are also chosen by comparing the meaningfulness of the topics found using different models (with different covariates). The analyst often interprets results by grouping the texts that show a prevalence of topics but the algorithm itself is not a clustering algorithm.

In other machine learning techniques, there needs to be a minimum amount of data to be able to fit the model. In STM, the number of topics need to be proportionate to the number of texts included but the length of the texts need not be long. A few sentences are usually enough to show an opinion. Roberts et al. (2014) use STM on survey responses. Short texts will be more likely to include only one topic. In our case, each text was at least half a page. Appendix 2 gives the number of words analyzed.

We refer to Blei et al. (2010), Grimmer and Stewart (2013), McFarland et al. (2013), Roberts et al. (2015b) for a complete description of the mathematical foundations, use and pitfalls of structural topic modeling. While this method allows for an inductive approach supporting discovery, it can also create correlations with no meaning at all. Following some steps can help avoid those and increase the robustness of the results. These include being careful and transparent about (1) cleaning the texts, (2) choosing the number of topics, (3) naming the topics and (4) validating the results (Grimmer and Stewart, 2013).

Preprocessing the text means getting rid of very frequent but non meaningful words, such as “and”, called stopwords. When doing it, the analyst needs to check that the resulting texts are made of meaningful words and that there still are a good amount of words in each document. Stemming is another preprocessing technique where words with the same root, such as “investor” “investment” and “investing”, are all grouped together as “invest-” as they carry very similar meanings.

Choosing the number of topics rests on finding topics that are cohesive, with high-probability words for the topic co-occurring within documents, and exclusive, meaning that the top words of the topic are unlikely to appear within top words for other topics (Mimno et al., 2011). However, topics will still always have some words in common that correspond to the overarching theme of the texts studied. Roberts et al., (2014) show two meaningful topics on immigration both with “illeg-” and “immigr-” among the top words, but the other ones show different perspectives on immigration. Quantitative measures helping to find a number of topics give a frontier and not a final solution. Ultimately, the final choice rests on the qualitative reading of the researcher (Chang et al., 2009). All topics won’t also necessarily be meaningful. The studies using STM have tended to use a very high number of topics (50 to 100 for Blei, 2012). One might choose 7 instead of 5 topics because it makes 5 topics very cohesive and exclusive, even though the 2 others are less interpretable, which is the case for our analysis of investment strategies. We show quantitative measures following the recommended methods for step 1 and 2 in Appendix 2.

To name the topics, we looked at the most frequent tokens appearing in each topic, and at graphs comparing the most frequent words between two topics (shown in Figure 3). We also read the full texts of the organizations that had the highest prevalence of each topic.

We validated the topics and groups of organizations proposed by the algorithm through our reading of annual reports, and by looking at the correlation between missions found by the algorithm and the categories of missions illustrated by Clark et al.'s case studies (2013) and described in those terms by PWC (2016: 6): “Capital maximization (Building a risk-adjusted capital base for the growth and preservation of national wealth); Economic development (Investment to boost a country’s long-term productivity); Stabilization (Macroeconomic management and economic smoothing). We do separate the mission “pension reserve” from the mission “capital maximization” when the organization is a pension fund and not a SWF (those two fall under capital maximization in the PWC analysis). Table 1 shows that we do find a correlation between using predominantly the topic that we named “national development”, and the organizations categorized as having a development mission in the literature. We also find a correlation between “stabilization” and the topic we named “global financial capitalism” to follow Monk’s (2011) and Clark and Monk's (2012a) analyses. This is sensible since stabilization often means investing passively on international markets. Table 2 does the same exercise for investment strategies and finds reassuring positive correlations between the strategy of managing risk and the organizations classified as PPFs, and the strategy of development and SWFs focused on development.

Table 1.

Correlation between topics in missions and typologies from the literature.^a

Topic	Global financial capitalism		Long-term investment		National development
Type of SI	Coeff	Pr(>\|t\|)	Coeff	Pr(>\|t\|)	Coeff	Pr(>\|t\|)
Public Pension Funds (PPF)	0.518	9.2E-08***	0.181	0.063	0.058	0.507
Sov. Dev. Funds (SDF)	–0.464	7.07E-04***	–0.131	0.380	0.798	9.51E-08***
Sov. Wealth Funds (SWF)	–0.386	6.07E-04***	0.123	0.302	0.213	0.039*
Missions
Capital maximization	0.234	0.026*	0.344	0.002**	0.182	0.077
Development	–0.173	0.178	–0.258	0.068	0.516	3.14E-04***
Pension Reserve	0.256	0.067	–0.118	0.430	–0.142	0.283
Stabilization	–0.228	0.126	0.019	0.907	0.082	0.616

To save space, we presented only significant correlations. The models tested are with all types of SI, or all types of missions as independent variables, and the topics proposed by the STM as dependent variables.Asterisks visually help to classify p values: *: P ≤ 0.05, **: P ≤ 0.01; ***: P ≤ 0.001

Table 2.

Correlation between topics in investment strategies and typologies from the literature.

Topic	Managing risk		Active management		Development
Type of SI	Coeff	Pr(>\|t\|)	Coeff	Pr(>\|t\|)	Coeff	Pr(>\|t\|)
PPF	0.288	0.0002***	0.251	0.0052**	0.009	0.878
SDF	–0.265	0.0224*	–0.149	0.2891	0.596	3.49E-06***
SWF	–0.209	0.0296*	–0.011	0.9186	0.116	0.172
Label	Maximizing returns on the market		Active management		Development
Missions	Coeff	Pr(>\|t\|)	Coeff	Pr(>\|t\|)	Coeff	Pr(>\|t\|)
Capital max.	0.291	0.0012**	0.339	0.0009***	0.016	0.8226
Development	–0.273	0.0154*	–0.194	0.1460	0.433	0.0001***
Pension Reserve	–0.131	0.2505	–0.123	0.3361	–0.01	0.9040
Stabilization	–0.034	0.8246	–0.158	0.3334	0.10	0.4349

These correlations help us validate some of the topics we found. For the topics that did not match, we qualitatively assessed the usefulness and meaning of new topics. We notably find that in the investment strategies, two topics are not as telling and meaningful as others, meaning they do not seem to be as consistent as others in the vocabulary used, and say therefore less about the organizations that use them.

Choosing external characteristics to test correlations with topics

A specificity of STM is that we can assess correlations between topics and external characteristics. We included regions as a factor, because geographical proximity might come with cultural and economic exchanges, leading to similar topics. We used the regions of the World Bank¹ and modified their list to separate countries from the “Pacific” from “East Asia”, which leads to the following 8 regions: North America; Central & Latin America; Europe & Central Asia; Middle East & North Africa; Africa; South Asia (no SI there); East Asia; Pacific.

Because SWFs are closely linked to their governments, we included political regime as a factor. We picked the democracy index from The Economist Intelligence Unit (2016) for several reasons. First, it uses criteria about the functioning of the political regimes as well as perceptions from the population through surveys, is transparent about the criteria used, and publishes a note about methods and reasons for scores with its results. It also provides a numerical score which is more precise for our calculation of correlations. Although it is not transparent about the experts who give the scores, we feel that there is enough information available in the report included in our source for readers to critique the value of these scores. The Economist stance is aligned with the values of global financial capitalism we have described, and we expect to see a correlation between the use of this themes and the degree of democracy given by this index.

However, we note that there is a correlation between regions and political regimes, with North America and Western Europe being on average full democracies, while the Middle East and North Africa are on average authoritarian regimes, and the rest are on average hybrid regimes. Figure 1 shows the distribution of our selection of long-term investors (SWFs and some PPFs) with and without an official report per type of political regime and per geography.

Figure 1.

Number of SWFs and PPFs (total and who published a 2015 report) by political regime type and by region.

We see that full democracies, Europe, North America, and Pacific regions overwhelmingly publish official reports, as expected from the fact that those reports are a norm from these regions and regimes.

Organizational characteristics of SWFs have also been underlined as potentially important. We include the size of assets under managements, year established and number of international associations the SWFs are a member of (Institutional Investors Roundtable, Long-Term Investors’ Club, IFSWF, Forum Mondial des Caisses des Depôts et Consignations). Membership in international associations could lead SWFs to adopt a common vocabulary per association, or to abide by a global set of values that supports their internationalization (M. Djelic and Quack, 2003b). In Figure 2, we see that organizations who are part of international associations also tend to publish an official report.

Figure 2.

Distribution of SWFs and PPFs per number of international organizations they are a member of.

Results from the topic modeling analysis of missions and investment strategies

An evolution of the missions of sovereign investors

As described in the section above, we find that five topics are significant to describe the missions of SWFs and pension funds. For each topic, the STM algorithm finds a list of words that appear most frequently together. Two topics often share frequent words. The difference is the frequency with which they appear with other words for each topic. Figure 3, generated by the STM package, illustrates this concept. The size of the stem word is proportional to its number of occurrences, while the distance between stem words represent how close they appear together in a topic.² The position on the vertical axis is random, but the horizontal axis shows if a stem word appears clearly more frequently in one topic (extreme left and right) or if it appears as much in the two topics (center position).

Figure 3.

Graph generated by the STM package to show the stem words comparatively characterizing two topics.

Figure 3 shows that the topic we named “global financial capitalism” emphasizes the words building on finance, assets, management, return, and risk, while, comparatively, the topic that we named “national development” features frequently together words related to development, the economy and companies.

To name each topic, we also read the mission statements of organizations that showed a prevalence of more than 90% for a topic. Below is an extract from the Canada Pension Plan Investment Board, a public pension fund that features the topic “global financial capitalism” with a prevalence superior to 90%: “The CPP Fund […] helps Canadians build financial security in retirement. The sole mandate of CPPIB is to invest the Fund ’s assets on behalf of the CPP’s 18 million contributors and beneficiaries to maximize returns over the long term without undue risk of loss .”

On the contrary, Saudi Arabia’s Public Investment Fund featured the topic of “national development” with a prevalence superior to 90%. Below is an extract from its mission statement: “[…] a leading economic player contributing to the development of national investment entities constituting the foundation for overall development while contributing to the establishment of many of the largest companies in Saudi Arabia […] The PIF has throughout the years financed many key projects and companies and provided financial support to projects of strategic importance to the national economy .”

Both organizations are investors and use the vocabulary of finance and investment. We don’t intend to downplay this aspect here, as all the organizations we study use predominantly finance and investment vocabularies. The difference lies in a certain preference for “invest” vs “fund”, which might indicate a more active vs passive stance, and the presence of other words next to those, which shows different ways to invest. We think the two extracts above clearly show two very different types of investment missions. The first one emphasizes a more classical, neoliberal view of finance, maximizing returns and minimizing risks, that characterizes global financial capitalism. The second one refers to investing in companies and projects that supports the development of the national economy, which can be linked to a refurbished state capitalism.

We identified a total of five missions compared to the four found in the literature. From comparisons of each pair generated by the STM package, we created the map in Figure 4 to show how all topics compare to each other. The size of the stem word shows its frequency in the topic, and stem words close to each other in topics appear together on the map. Words on the lines between two topics appear as frequently in both topics.

Figure 4.

Stem words characterizing each topic in mission statements as compared to other topics.

Compared to the missions identified in the literature, we find more nuance in the stances of the SWFs. For example, we do not find a mission of “capital maximization” per se as described in the literature. Instead, capital maximization is declined through the prism of “global financial capitalism”, emphasizing risks, returns and the markets; and through the prism of “storing national wealth”, in which case “fund” and “nation” appear with risks and returns. We find a vocabulary linked to the mission of “stabilization” of the national economy, meaning protecting it from changes in commodity prices (Dixon and Monk, 2014). We add that organizations that use it also often underline their link to their national administration. Finally, we find a new mission compared to existing typologies, that of “long-term market investments”, which emphasizes the long-term investment horizon of the organization.

In addition to identifying topics, the STM results gives us the prevalence of topics used by each organization to describe their missions. We use this result to map SWFs that use predominantly one topic in Figure 5.

Figure 5.

Map of SWFs and PPFs based on the predominance of topics in their mission statements.

In Figure 5, organizations with a clearly prevalent topic are at the edge of the figure. STM is a mixed-membership model, but the majority of the text we analywe feature only one topic. Those that blend two topics are at the frontier between those two topics. Organizations close to the center such as PGGM have 2 or 3 topics with the same prevalence, and we place them in the wedge of the topic with the most prevalence. We acknowledge that their interpretation is harder, but there are only few of them.

We see that few organizations predominantly use topics linked to storing national wealth and stabilizing their economies (topics 1 and 5), but that those organizations have a large amount of assets under management. This makes sense as they are often linked to their countries’ central banks. Six SWFs, a majority of which are small and from democracies, predominantly use the long-term market investments mission. As does the American PPF TIAA-CREF. The two most popular topics are global financial capitalism and national development. Eight organizations feature global financial capitalism in more than 95% of their mission statements. All of them are democracies and 6 of them are PPFs. 13 SWFs feature national development in more than 95% of their mission statements. A majority of those are small and from authoritarian countries.

A group of six large SWFs and one large PPF blend the topics of national development and global financial capitalism. This is surprising given the Santiago Principles’ push for SWFs to create distance with their governments’ missions and look solely for financial returns. The topic that we name “national development” recognizes the role of the SWF to support the development of its national economy and brings a focus on investment in companies and projects, instead of just market shares. The SWFs that use both topics are prominent in the literature, and this duality has often been underlined as core to SWFs (Clark and Monk, 2012; Dixon and Monk, 2011; Monk, 2011). This map shows that using the values and tools of global financial capitalism at the service of national development is particularly striking for this group of well-known SWFs, that have also been part of large international negotiations and foreign investments. However, many other SWFs position themselves more toward development or more toward what we called long-term market investments, and this blend of local and global norms is not the norm for all SWFs.

We also tested the correlation between use of topics and geographical, political and organizational characteristics of the SWFs. Table 3 shows these correlations for the three topics that show significant correlations.

Table 3.

Correlations between external characteristics and topics used to describe mission statements.

	Topic 3: Global Financial Capitalism		Topic 4: Long-term Market Investment		Topic 2: National Development
	Coeff. Estimate	Wald Pr(>\|z\|)	Coeff. Estimate	Wald Pr(>\|z\|)	Coeff. Estimate	Wald: Pr(>\|z\|)
Model 1 – Regions
Central Asia	0.001365	0.99136	0.20492	0.16419	0.232351	0.13513
East Asia	0.183446	0.05924	0.12219	0.20214	0.291479	0.00533**
Europe	0.172273	0.20125	0.24529	0.10197	0.207804	0.13047
Latin America & Caribbean	0.009306	0.966	0.82649	0.00276**	0.004953	0.98347
Middle East & N. Africa	0.08745	0.38829	0.03237	0.74603	0.748415	6.87E-08***
North America	0.589763	9.64E-07***	0.3263	0.00292**	0.051627	0.62272
Pacific	0.532705	0.00804**	0.44783	0.03225*	0.003776	0.98204
Sub-saharan Africa	0.139017	0.49365	0.23128	0.34402	0.495285	0.05763
Model 2 – Number of international clubs + OECD + Year of creation + log(AUM)
(Intercept)	–3.70983	0.26193	1.244747	0.7159	6.662666	0.07097
# of international clubs	–0.01647	0.79751	–0.03856	0.589	0.056443	0.42567
OECD	0.342688	0.00047***	0.187994	0.0722	–0.42066	0.00016***
Year	0.001843	0.26133	–0.00047	0.7844	–0.00303	0.09828
log(AUM)	0.038621	0.09882	–0.03705	0.1705	–0.04918	0.0671
Model 3 – Democracy Score + Year of creation + log(AUM)
(Intercept)	–4.82427	0.100083	2.46E-01	0.9364	8.371403	0.00973**
Democracy Score	0.06869	0.000121***	4.45E-02	0.0234*	–0.08329	4.70E-05***
Year	0.002247	0.123686	–6.99E-05	0.9635	–0.00371	0.02085*
log(AUM)	0.044006	0.066721	–4.10E-02	0.1234	–0.04364	0.0971
Model 4 – international club membership IFSWF + FMCD + IIR + LTIC
(Intercept)	0.25261	0.00232**	0.19517	0.0172*	0.28851	0.00242**
IFSWF	–0.15447	0.14716	0.13257	0.2155	0.03312	0.78706
FMCD	–0.17394	0.53998	–0.11478	0.7071	0.28089	0.41197
IIR	0.29176	0.01803*	–0.14141	0.2205	–0.03603	0.79569
LTIC	–0.06848	0.71598	0.06882	0.733	–0.04978	0.82399

Model 1 shows that organizations that use the topic of global financial capitalism and long-term investing to describe their missions are significantly from North America and the Pacific region. On the contrary, organizations using the topic of national development are significantly from the Middle East & North Africa and from East Asia.

Model 2 and 3 show that organizations using the topic of global financial capitalism tend to be bigger, and that organizations that use the topic of national development tend to be smaller and created more recently.

They also show that being part of the OECD and from a democratic country positively correlates with using the topic of global financial capitalism, and negatively correlates with using the topic of national development, which is in line with the idea that the international community and western countries supports the values and vocabulary of global financial capitalism, and do not mix in developmental goals. Organizations not from the OECD and with low democracy scores correlate with the expression of national development goals in their mission statements. This result tends to show that refurbished capitalism ideals are legitimate enough for some SWFs to declare them in official reports but cannot be said to have a global aura just yet.

Finally, organizations who are part of the international professional association the International Investors Roundtable (IIR), despite coming from a variety of regions and countries, also use predominantly the topic of global financial capitalism, as it includes many of the PPFs that are at the forefront of global financial capitalism and of the large SWFs that we categorized as hybrid. We now turn to the analysis of investment strategies. We find that long-term market investment is indeed a prominent goal for SWFs, but not that it is correlated with international associations. Instead, it seems to be a way for SWFs from democracies to differentiate their goals from PPFs, especially in the Pacific region and Canada.

Topics in investment strategies

Investment strategies are a little harder to study using our method because we focus solely on words and strip the texts out of numbers. We therefore do not claim any detailed and nuanced understanding of each organization’s financial stance here. Our goal is to see if some broad trends are visible among SWFs, and if the clusters of SWFs found in missions correspond to the same clusters in terms of their investment strategies. The topics we find below are not all coherent and telling of a really differentiated strategy. Indeed, we find that topic 5 and 6 gather common words to describe investment strategies but do not really tell us anything meaningful. However, picking 7 topics helps to create 5 cohesive and meaningful topics.

Some topics show that differences identified in the missions also appear in the investment strategies. This is particularly the case for the topic of “development”, as opposed to “maximizing returns on the market” and “active management”. Figure 6 shows the stem words characterizing the seven topics found for investment strategies, while Figure 7 maps SWFs and PPFs according to the prevalence of topics they use. Organizations at the edge of the figure use one topic in their strategies while the ones toward the middle use a mix of topics and are in the wedge of their most prevalent topic.

Figure 6.

Stem words characterizing topics in investment strategies.

Figure 7.

Map of SWFs and PPFs per the prevalence of topics used in their investment strategies.

First, all the organizations categorized as using a “development” topic to describe their investment strategy also featured “development” predominantly in their missions. These organizations describe investment in projects and companies that support their national economies. This quote from Mubadala illustrates this position: “Investment approach: Mubadala’s strategy is built on the creation of partnerships and on long-term, capital-intensive investments that deliver strong financial returns, contribute to the growth and diversification of Abu Dhabi’s economy, and create opportunities for current and future generations in the United Arab Emirates.”

RDIF also emphasizes projects, companies and co-investment in the Russian economy: “RDIF has invested in projects across more than 95% of Russia’s regions . RDIF, together with its partners, invest in leading companies in the fastest-growing sectors of the Russian economy .”

A final example is found in Khazanah: “We partner our investee companies in collaborative manner to venture into new markets and work with public and private sector entities to help catalyse the growth and development of Malaysia.”

However, not all the organizations who declared a development mission talk about “national development” in their investment strategy. This shows that these organizations that had a predominant mission of development do not necessarily invest directly in companies and projects for national development, which means they also fulfill this goal by investing in shares on international markets, as described by Dixon and Monk (2016).

On the contrary, organizations mapped onto the topic of “Maximizing returns on the market” are characterized by a detailed description of the performance of each asset class compared to a benchmark portfolio, as these extracts from Norges Bank and the Alaska Permanent Fund illustrate:

“The fund’s equity investments returned 3.8 percent in 2015 after gains on European and US stocks. The fund’s investments in developed markets returned 4.9 percent, while investments in emerging markets returned -7.0 percent.” (Norges);

“For fiscal year 2015 the total public stock portfolio declined 0.1 percent, underperforming the MSCI ACWI IMI benchmark, which returned a positive 0.8 percent.” (APF).

Organizations classified as “Active Management” also detail the performance per asset class but they most often also describe an active and differentiated strategy, most notably a split of the portfolio between index funds, and an actively managed portfolio that aims at taking advantage of their long-term horizon, as the extracts from GIC (Singapore), the Future Fund (Australia) and ADIA (Abu Dhabi) below show:

“For a given level of risk , we strive to optimise the GIC Portfolio to achieve the best possible long-term return for a variety of economic scenarios. […] We separate active skill-based “alpha” strategies from beta activities at GIC to enable greater effectiveness and efficiency.”

“Within each category, [The Future Fund] develops a fine-grained investment strategy that is consistent with our core objectives and approach to total portfolio construction. […] We aim to build portfolios that are as resilient to the investment environment as is practicable and efficient, so we strive for genuine diversification that achieves greater balance in portfolio construction.”

“In order to achieve its long-term objectives , ADIA must be able to execute on its desired asset allocation in a timely fashion, in size, while minimising transaction costs. It is for this reason that slightly more than half of ADIA’s portfolio consists of index-replicating, or passive, strategies within quoted markets. This is offset by skillfully designed, actively managed investments across asset classes, in areas with genuine potential to generate market outperformance, or alpha, over the long-term .”

Table 4 shows the significant correlations between external characteristics and the topics found in SWFs and PPFs’ investment strategies.

Table 4.

Significant correlations between SWFs and PPFs’ characteristics and topics in investment strategies.

	Topic 2: Maximizing Returns on the Market		Topic 3: Active Management		Topic 4: Development
	Coeff. Estimate	Wald Pr(>\|z\|)	Coeff. Estimate	Wald Pr(>\|z\|)	Coeff. Estimate	Wald: Pr(>\|z\|)
Model 1 – Regions
Central Asia	0.089	0.481377	0.16288	0.25501	0.229875	0.0786
East Asia	0.11142	0.179091	0.240332	0.01818*	0.140181	0.0938
Europe	0.1342	0.235166	0.061311	0.62733	0.046501	0.6909
Latin America & Caribbean	0.24017	0.36674	0.001985	0.99308	0.002578	0.9897
Middle East & N. Africa	0.0296	0.731732	2.95E-01	0.00725**	0.505507	5.96E-06***
North America	4.51E-01	0.000185***	0.18552	0.10411	0.005442	0.9492
Pacific	0.04311	0.767589	0.513944	0.00801**	–0.0009	0.9949
Sub-saharan Africa	0.01783	0.911279	0.171749	0.42804	0.091835	0.6041
Model 2 – Number of international clubs + OECD + Year of creation + log(AUM)
(Intercept)	–1.448	0.611	–4.95293	0.101	–0.07702	0.97988
# of international clubs	–0.00529	0.928	0.082008	0.193	0.017798	0.75456
OECD	0.177012	0.057	–0.03513	0.716	–0.24888	0.00608**
log(AUM)	0.004269	0.843	0.026336	0.281	–0.04173	0.06272
Year	0.000765	0.589	0.002516	0.094	0.000248	0.87037
Model 3 – Democracy Score + Year of creation + log(AUM)
(Intercept)	–1.87719	0.4747	–3.70786	0.22	0.170729	0.9517
Democracy Score	0.041314	0.0139*	0.0007	0.97	–0.05235	0.0023**
Year	0.000886	0.4964	0.001912	0.203	0.000236	0.8662
log(AUM)	0.003066	0.8836	0.029496	0.24	–0.04048	0.0715
Model 4 – IFSWF + FMCD + IIR + LTIC
(Intercept)	0.16084	0.0253*	0.12434	0.115	0.17668	0.0303*
IFSWF	0.01061	0.9121	0.13911	0.171	0.01634	0.8735
FMCD	0.23724	0.347	–0.22083	0.384	–0.05697	0.8397
IIR	–0.02627	0.803	0.25299	0.022*	–0.08109	0.4625
LTIC	–0.11304	0.5134	–0.1127	0.524	0.03051	0.8773

Model 1 shows that organizations from North America predominantly use the topic of “Maximizing returns using the market”. Organizations from the Pacific region, East Asia and to a lesser extent the Middle East & North Africa use the topic of “Active management” to describe their investment strategies. Finally, organizations featuring the “Development” topic are significantly positively correlated with the Middle East region, but also with Central and East Asia with less certainty.

In line with the correlations found for mission statements, Models 2 and 3 show that organizations from the OECD and democratic countries use the topic of “Maximizing returns on the market”, describing investments in details, and tend not to insist on national investments to support their economies, while organizations from authoritarian countries do the contrary. Organizations claiming investments in their national economies also tend to be smaller.

Finally, being a member of the IIR correlates with the topic of “active management”. The IIR has regular meetings and conferences in which members think about new ways of investing and to take advantage of their long-term horizons, and members also underline this aspect in their official reporting. The members of this group are large SWFs and PPFs, older, and interested in both domestic and international investments.

This analysis has shown that in SWFs and PPFs’ official reports, we were able to identify 5 meaningful topics in their missions as well as 5 meaningful topics in their investment strategies. SWFs did not necessarily use only one topic to describe their missions and strategies, most notably, we found that a group of SWFs of large size and from various countries seemed to blend missions of development with the vocabulary of global financial capitalism. We found that SWFs with development missions are mostly consistent and differentiated in this goal in their investment strategies. A strategy we called “Active Management”, which often describes a mix of market investments and direct investments according to an investment vision, is particularly popular, notably among the large SWFs that are part of the IIR and featured missions combining the topics development and global financial capitalism. In terms of influencing external characteristics, we found that geography and political regime are still very much linked to two visions of capitalism, one in line with neoliberal finance for western democracies and one aligned with state capitalism for authoritarian regimes of East Asia and the Middle East. The international professional associations we posited could lead to a new identity do not prevail over these more traditional influences, but they seem to open the door to the emergence of investment strategies that are more innovative and active than traditional market investments. This indicates that through international clubs, SWFs could learn about new investment tools and methods, learn from each other’s investment experience and slowly develop models of investment (Monk et al., 2017), without changing their missions and values, which would still be dictated by each national government.

Discussion

The analysis above gives an updated vision of the topics used by SWFs to describe their missions and investment strategies and shows where each SWF and PPF lie on a cultural map of missions and investment strategies.

While SWFs and PPFs had been described as long-term investors (OECD, 2013), this orientation had not been identified as a prevalent topic to describe a mission among some SWFs and PPFs in the same way development or stabilization had been in past typologies (Dixon and Monk, 2014). Here, we were able to show that although all SWFs and PPFs could claim their long-term horizon, only a selected group, predominantly SWFs from western democracies, could be identified based on this mission. For development or stabilization SWFs, long-term horizon appears to be more at the service of other values. We also posit that this topic helps differentiate SWFs from PPFs who use a more conservative neoliberal stance in western democracies. These findings brings a perspective to call for both PPFs and SWFs to become long-term investors and confirms views that important differences in missions and stances can prevent PPFs and SWFs to massively join forces under similar long-term investing approaches (OECD, 2014; Warren, 2014).

We also identified a group of large SWFs with hybrid goals, underlining both their will to optimize their portfolios for risks and returns, and their mission to support the development of their national economy. These SWFs are often described in the literature on SWFs, notably to describe how they differ from other institutional investors, manage international and local investments, and could spearhead the emergence of a new type of investor (Bachher et al., 2015; Clark et al., 2013; Dixon and Monk, 2012; Gelb et al., 2014 ). Our results show that their situation should not hide the fact that SWFs are still widely diverse and differentiated, with some clearly development-oriented, some clearly financial-return and market oriented, and some still close to state entities in charge of stabilization and the administration of national wealth.

In the variety of missions and investment strategies, and the correlation between some topics and some regions and political systems, we do find that SWFs are portraying how the global, rational ideal of financial capitalism is adapted when meeting national, cultural and historical realities. We think our map are aligned with that view and advance it by illustrating several paths of adaptation of capitalism through SWFs. We show that there is a place for classical, traditional global financial capitalism, especially in countries where it is the norm and for organizations that are older and risk-averse such as pension funds. In addition to the predicted adaptation of capitalism for national protection through stabilization, reserve and even development, we also see an adaptation of it among SWFs of western democracies toward investing for the long-term, which might be a sign of bending some neoliberal investment rules and tools.

We also see a strong presence in official reports, and therefore a renewed legitimacy of a refurbished state capitalism as we described in the literature review, recognizing the advantages of using capitalism at the service of national development. In fact, we see this as the main goal of many newly created small SWFs in developing countries. SWFs could be an ideal instrument to push this stance. This echoes the interest in the literature for sovereign development funds and strategic funds (Clark and Monk, 2015; Haberly, 2011; Halland et al., 2016). We posit that one could see a positive evolution toward the use of the development topic from 2007 to today if one were to do the same study over several years. We could see if more and more countries seem to publicly adopt this stance, and if it is gaining ground among western organizations as well. We would be curious to see if the date of creation and the region of SWFs is highly correlated with the use of the development topic over time.

Finally, belonging to the Institutional Investment Roundtable was correlated with an investment strategy that we called “active management”, that corresponds to taking an active stance in defining a unique strategy for the fund, pioneering new tools to manage the specific long-term risks of SWFs; and trying to find new asset classes that would be fitted to long-term horizons and deep pockets. The active management stance cut across geographies, political regimes, and being a SWF or a PPF. We therefore see it as the premise of a potential professionalization of SWFs as a community that brings together organizations that can pioneer new ways of investing and give itself new rules in that matter. However, it seems to show that this group would stay away from defining a common identity, mission or even common values, as those are strongly influenced by regional cultures and political systems. In this case, international clubs could either stay networking events where SWFs could find occasional partners and deals; or they might establish best practices, collaborate with academics forging new investment theories, and be a cradle for these to spread among SWFs across frontiers.

Conclusion

We started by identifying that SWFs could mirror several institutional processes. They portray variegated capitalism, illustrating how capitalism is adapted in different regions and political systems. They also show how a refurbished state capitalism is emerging and maybe gaining ground beyond historically state capitalist countries. Finally, we posited that the creation of new international professional organizations of SWFs and PPFs could lead to new missions and investment strategies emerging out of an emergent institutionalization process, independently of local cultures and political pressures.

This study has used a novel methodological approach, structural topic modeling, on a uniquely large and global set of 40 SWFs’ and 17 PPFs’ official reports, to show how SWFs as a community portrayed these different influences, and reacted to institutional influences from the global and national levels, both political or professional.

The result is a visual structuration of the landscape of SWFs and PPFs that helps to see similarities between some of these organizations and to understand the variety of the missions and investment strategies of all these organizations. We were able to show that geography and political regimes explain a lot of these similarities, helping to show one pole of neoliberalist values, versus a pole with a refurbished state capitalism stance. But, we also showed some middle grounds, SWFs that evenly mixed these two seemingly opposite value systems, as well as new grounds, SWFs that talked about long-term investing in their missions, and SWFs and PPFs that proposed new ways of investing to match their unique identity as long-term investors.

In doing so, we aimed to show the potential of structural topic modeling as a new method for economic geography and notably to contribute to the use of the concept of variegated capitalism (Peck and Theodore, 2007). Structural topic modeling allows researchers to study large amounts of qualitative data faster, to look at their data through a new lens as they compare texts that share topics revealed by the algorithm, and to have a quantitative measure of the correlation between the occurrence of a topic in texts and meta-data about those texts, such as the countries and regions of the authors of those texts. The increasing availability in a digital form of institutional communication means that we now have access to a treasure trove of written communication of entire communities over time. In sociology and political science, STM is increasingly used to analyze these vast corpora of texts to reliably identify patterns of language, and to identify the factors that generate those patterns of language differentiation (McFarland et al., 2013: 609). It can do the same in economic geography to help qualify how the capitalist ideology evolves among different organizations and regions.

We think our study could be complemented by using STM to study annual reports of SWFs over time to see influences between organizations and an evolution of the groups we have identified, in a similar way Rockmore et al. (2016) have done when studying constitutions over time around the world. Finally, we think mapping out the influence of professional organizations on top of more traditional institutions related to markets and the State will help identify another avenue of change and adaptation of the capitalist system.

Footnotes

Acknowledgements

The authors wish to thank three anonymous reviewers, Prof. William Barnett, Prof. Justin Grimmer, Prof. Raymond Levitt, Prof. W. Richard Scott, Prof. Vedran Zerjav, Bertrand Decoster, the members of the Stanford Global Projects Center and the participants in the 2017 Engineering Project Organization Conference for useful comments on previous versions of this paper. None of the above is responsible for any errors or omissions.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding from the Stanford Global Projects Center supported this research.

Notes

References

Albert

Whetten

(1985) Organizational identity. Research in Organizational Behavior 7: 263–295.

Argenti P and Druckenmiller B (2004) Reputation and the Corporate Brand. Corp Reputation Rev 6: 368–374.

Bachher JS, Clark GL, Monk A and Sridhar K (2014) “The Valley of Opportunity”: Rethinking Venture Capital for Long-Term Institutional Investors. Alternative Investment Analyst Review 3(1): 63–75.

Bachher

Monk

AHB

(2013) Platforms and vehicles for institutional co-investing. Rotman International Journal of Pension Management 6(1): 64–71.

Backer

(2009) Sovereign wealth funds as regulatory chameleons: the Norwegian sovereign wealth funds and public global governance through private global investment. Georgetown Journal of International Law 41(2): 425.

Balding

(2009) Who’s afraid of sovereign wealth funds? Revue D'économie Financière (English ed.) 9(1): 201–210.

Blei D, Carin L and Dunson D (2010) Probabilistic topic models. IEEE Signal Processing Magazine 27(6): 55–65. https://doi.org/10.1109/MSP.2010.938079

Chang

Y-C

Yang

P Y

Chen

M-H

(2009) The determinants of academic research commercial performance: Towards an organizational ambidexterity perspective. Research Policy 38(6): 936–946.

Clark

Dixon

Monk

(2013) Sovereign Wealth Funds, Legitimacy, Governance and Global Power. Princeton: Princeton University Press.

10.

Clark

Monk

(2012a) Modernity, imitation, and performance: Sovereign funds in the Gulf. Business and Politics 14(1): 1–17.

11.

Clark GL and Monk AHB (2012b) Principles and Policies for In-House Asset Management. The Journal of Financial Perspectives 1(3): 1–14.

12.

Clark Peter B and Ashby Monk (2015) “Sovereign Development Funds: Designing High-Performance, Strategic Investment Institutions.” Working Paper, Stanford University, Stanford, California.

13.

Dixon

(2014) Oxford Scholarship Online Finance, and Society Variegated Capitalism and the Firm, (January 2019), 1–24. Print ISBN-13: 9780199668236. Published to Oxford Scholarship Online: August 2014. DOI: 10.1093/acprof:oso/9780199668236.001.0001

14.

Dixon

Monk

AHB

(2011a) Rethinking the sovereign in sovereign wealth funds. Transactions of the Institute of British Geographers 37(1): 104–117.

15.

Dixon

Monk

AHB

(2011b) The Design and Governance of Sovereign Wealth Funds: Principles & Practices for Resource Revenue Management ∗, working paper, University of Bristol. 1–15.

16.

Dixon AD and Monk AHB (2012) Reconciling transparency and long-term investing within sovereign funds. Journal of Sustainable Finance & Investment 2 (3-4): 275–286. DOI: 10.1080/20430795.2012.703126

17.

Dixon

Monk

AHB

(2014) Financializing development: Toward a sympathetic critique of sovereign development funds. Journal of Sustainable Finance & Investment 4(4): 357–371.

18.

Dixon AD and Monk AHB (2017) A Simple Typology of Sovereign Development Funds. In M. Rietveld and P. Toledano (Eds.), The New Frontiers of Sovereign Investment. Columbia University, p. 83.

19.

Djelic

Quack

(2003a) Conclusion: Globalization as a double process of institutional change and institution building. In: Globalization and Institutions: Redefining the Rules of the Economic Game. pp. 302. editors: Djelic, M & Quack, S. DOI: 10.4337/9781781950012.00001

20.

Djelic

Quack

(2003b) Theoretical Building Blocks for a Research Agenda Linking Globalization and Institutions Towards a Typology of Institutional Arguments. In M. Djelic and S. Quack (Eds.) Globalization and Institutions: Redefining the Rules of the Economic Game. Chapter 1, p. 15.

21.

Djelic

Quack

(2007) Overcoming path dependency: Path generation in open systems. Theory and Society 36(2): 161–186. http://doi.org/10.1007/s11186-007-9026-0

22.

Drezner

D W

(2008) Sovereign wealth funds and the (in)security of global finance. Journal of International Affairs 62(1): 115–130.

23.

Esrock SL and Leichty GB (1998) Social responsibility and corporate web pages: Self-presentation or agenda-setting? Public Relations Review 24(3): 305–319.

24.

Fligstein N and McAdam D (2012) Change and Stability in Strategic Action Fields. A theory of Fields. Oxford: Oxford University Press. https://doi.org/10.1093/acprof:oso/9780199859948.003.0004

25.

G20/OECD (2013). G20/OECD High-Level Principles on Long-Term Investment Financing by Institutional Investors. OECD Publishing.

26.

Gelb

Tordo

Halland

, et al. (2014) Sovereign Wealth Funds and Long-Term Development Finance Risks and Opportunities. Policy research working paper. The World Bank.

27.

Greenwood R, Suddaby R and Hinings CR (2002) Theorizing change: The role of professional associations in the transformation of institutionalized fields. The Academy of Management Journal 45(1): 58–80.

28.

Grimmer

Stewart

B M

(2013) Text as data: The promise and pitfalls of automatic content analysis methods for political texts. Political Analysis 21(3): 267–297. http://doi.org/10.1093/pan/mps028

29.

Haberly

(2011) Strategic sovereign wealth fund investment and the new alliance capitalism: A network mapping investigation. Environment and Planning A: Economy and Space 43(8): 1833–1852. http://doi.org/10.1068/a43324

30.

Haberly

Wójcik

(2017) Earth incorporated: Centralization and variegation in the global company network. Economic Geography 93(3): 241–266.

31.

Hall

Soskice

(2001) An introduction to varieties of capitalism. In Hall

Soskice

(eds) Varieties of Capitalism: The Institutional Foundations of Comparative Advantage. Oxford: Oxford University Press, p.540.

32.

Halland

Noel

Tordo

, et al. (2016) Strategic Investment Funds, Opportunities and challenges. Policy Research Working Paper, The World Bank.

33.

Khaire

Durand

Jones

, et al. (2010) Changing landscapes: The construction of meaning and value in a new market category – Modern Indian art. Academy of Management Journal 53(6): 1281–1304.

34.

Kleiner

(2003) Building up an asset management industry: Forays of an Anglo-Saxon logic into the French business system. In: Djelic

Quack

(eds) Globalization and Institutions. Edward Elgar Publishing.

35.

Leuthesser L and Kohli C (1997) Corporate identity: The role of mission statements. Business Horizons 40(3): 59–66.

36.

Lucas C, Nielsen RA, Roberts ME, Stewart BM, Storer A and Tingley D (2012) Online Appendix: Additional Text Preparation Details and A Review for Text Analysis Techniques for Comparativists. Political Analysis 23(2): 1–23.

37.

Lucas C, Nielsen RA, Roberts ME, Stewart BM, Storer A and Tingley D (2015) Computer-assisted text analysis for comparative politics. Political Analysis 23(2). https://doi.org/10.1093/pan/mpu019

38.

Lucas

Nielsen

Roberts

, et al. (2015) Computer-assisted text analysis for comparative politics. Political Analysis 23(2): 254–277.

39.

Lucas

Nielsen

Roberts

, et al. (2015) Online appendix: Additional text preparation details and a review for text analysis techniques for comparativists. Political Analysis 23(2): 254–223.

40.

McFarland

Ramage

Chuang

, et al. (2013) Differentiating language usage through topic models. Poetics 41(6): 607–625.

41.

McNally

(2013) The Challenge of Refurbished State Capitalism: Implications for the Global Political Economic Order.

42.

Mehrpouya

(2015) Instituting a transnational accountability regime: the case of sovereign wealth funds and “GAPP”. Accounting, Organizations and Society 44: 15–36.

43.

Meyer

Pope

Isaacson

(2013) Legitimating the transnational corporation in a stateless world society. Legitimating the Transnational Corporation in a Stateless World Society 1(2010): 1–55.

44.

Mimno

Wallach

H M

Talley

, et al. (2011). Optimizing semantic coherence in topic models. Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing 2: 262–272.

45.

Monk

(2009) Recasting the sovereign wealth fund debate: Trust, legitimacy, and governance . New Political Economy 14(4): 451–468.

46.

Monk

(2011) Sovereignty in the era of global capitalism: the rise of sovereign wealth funds and the power of finance. Environment and Planning A: Economy and Space 43(8): 1813–1832. http://doi.org/10.1068/a43326

47.

Monk

Sharma

Sinclair

D L

(2017) Reframing Finance New Models of Long-Term Investment Management. Stanford University Press.

48.

OECD (2013). The Role of Banks, Equity Markets and Institutional Investors in Long-Term Financing for Growth and.

49.

OECD (2014). Pooling of Institutional Investors Capital – Selected Case Studies in Unlisted Equity.

50.

Papaioannou

M G

Park

Pihlman

, MPapaioannou@imf.orget al. (2013) Procyclical behavior of institutional investors during the recent financial crisis: Causes, impacts, and challenges. IMF Working Papers 13(193): 1.

51.

Peck

Theodore

(2007) Variegated capitalism. Progress in Human Geography 31(6): 731–772.

52.

PWC (2016). Sovereign Investors 2020 A Growing Force.

53.

Roberts

Stewart

Tingley

(2015a) STM: R package for structural topic models. Journal of Statistical Software 91(2).

54.

Roberts

Stewart

Tingley

(2015b) STM: R package for structural topic models. Journal of Statistical Software 91(2).

55.

Roberts

Stewart

Tingley

, et al. (2014) Structural topic models for open-ended survey responses. American Journal of Political Science 58(4): 1064–1082.

56.

Rockmore

Fang

Foti

N J

, et al. (2016) The Cultural Evolution of National Constitutions. Working paper.

57.

Scott

(2008) Approaching adulthood: The maturing of institutional theory. Theory and Society 37(5): 427–442. http://doi.org/10.1007/s11186-008-9067-z

58.

Tainio

Pulkkinen

Huolman

(2003) Global investors meet local managers: Shareholder value in the Finnish context. In Djelic

M-L

Quack

(eds), Globalization and Institutions. Edward Elgar Publishing.

59.

The Economist Intelligence Unit (2016). Democracy Index 2015 Democracy in an age of anxiety. Research report.

60.

Truman

E M

(2010) Sovereign Wealth Funds: Threat or Salvation?. Washington, DC.

61.

Wallach

Murray

Salakhutdinov

, et al. (2009) Evaluation methods for topic models. International Conference on Machine Learning 1: 1–3.

62.

Warren

(2014) Long-Term Investing: An Insitutional Investor Perspective. Research Report. Center of International Finance and Regulation.