The UN General Assembly Sponsorship Dataset
The UN General Assembly Sponsorship Dataset monitors co-sponsorship of draft resolutions at the UN General Assembly (UNGA) for all 193 members states (plus Palestine) from 2009 to 2020. In total, it has information on 3,937 L-Documents that form the basis of later resolutions. As multiple L-Documents can refer to the same resolution proposal (a root L-Document, followed by revisions or addenda), 2,757 “sequenced drafts” are also made available. The dataset is based on raw data retrieved from the UN Digital Library. It has a total of 28 variables related to drafts metadata, including date, title, originating committee, agenda item, participating groups, related documents, etc.) More importantly, it contains 194 variables on sponsorship by each of the UNGA member states, which allows to identify participating countries for every draft. The dataset offers both simple and more granular measures of country participation. Simple measures consist of binary scores indicating whether or not a country co-sponsored a draft. But because not all drafts are equally relevant for sponsoring states, two additional indicators are introduced to properly weigh country engagement: priority and ownership. These allow to identify which drafts were more attuned to state preferences, based on how soon countries joined and on the extension of their co-sponsors.
Figure 1: Two sample draft resolutions, their sequence of constitutive L-Documents, sponsoring countries, and attribution of priority and ownership scores.
Over the years, research on the UNGA has privileged analyses on voting patterns as a measure of interstate affinity. Yet, features inherent to the voting process impose significant limitations: the largest share of UNGA output is never put to a vote, the share that does get voted is disproportionally concentrated on disputed topics, and voting is just the endpoint of a long and complex process. As such, inferences on state preferences based on this metric can be prone to biases. Sponsorship of draft resolutions, on the other hand, can prove a more comprehensive indicator to outline national profiles, since it covers the unfiltered totality of countries’ production at the UNGA; like voting, the decision to sponsor is a form of signaling that can reveal state preferences.
What can you do with it?
The UN General Assembly Sponsorship Dataset can be used similarly to other legislator-decision datasets. Possible applications include:
(1) Country and topic profiling: by observing the topics of the drafts that delegations endorse, it is possible to derive the thematic profile of each member state. This allows to identify who were the “banner carriers” that consistently tabled propositions on a given topic, or list the preferred themes of specific countries and groups.
Figure 2: Absolute priority and ownership of UNGA drafts for a sample of countries. Black dots indicate drafts containing keywords intuitively regarded as relevant for each country.
(2) Coalitions and blocs: data can be used for network analysis, in which countries are considered nodes and drafts the links through which they cooperate. For example, community detection algorithms can be used to identify new coalitions at the UNGA based on the co-sponsorship network (Seabra and Mesquita 2022). Other network-based applications can entail calculating centrality scores to identify, for instance, countries playing brokering roles.
Figure 3: Communities detected in the co-sponsorship network.
(3) Dimensions and scaling: Dimensionality reduction techniques, such as PCA or spatial models for ideal point estimation, can be used to extract meaningful dimensions from total interactions between states. Such dimensions can reveal broad patterns of conflict and cooperation across all UNGA activities, or more focused estimates, if calculated for narrower committees or topics of interest.
(4) Interstate influence and other roles in bargaining dynamics: key research questions in UNGA literature on interstate influence, such as vote-buying or socialization effects, can be further explored. The drafting process, in particular, offers micro-level observations that are denser than yearly aggregates of voting scores, making them more reliable to trace observable behavior back to preferences. For example, priority scores can be used to map the precise moment when great powers endorsed a draft and if this signal increased the likelihood of co-sponsorship by ODA-receiving countries (Seabra and Mesquita 2022).
(5) Determinants of action: lastly, data also allow for causal research designs on country behavior or legislative outcomes. Examples would include modelling the determinants of a country’s decision to join a draft, survival models on how long until they adhere, or what factors explain the success of failure of a proposal, in case draft outcome is treated as the dependent variable.
Seabra, Pedro; Mesquita, Rafael, 2022, “UN General Assembly Sponsorship Dataset”, https://doi.org/10.7910/DVN/MPQUE2, Harvard Dataverse, V2
Seabra, Pedro and Rafael Mesquita, Beyond Roll-Call Voting: Sponsorship Dynamics at the UN General Assembly, International Studies Quarterly, Volume 66, Issue 2, 2022, sqac008, https://doi.org/10.1093/isq/sqac008