Measuring the dynamical evolution of the United States lobbying network

Abstract

Lobbying networks constitute complex political systems that mobilize vast human and financial resources to influence governmental decision-making, often with profound national and global consequences. A comprehensive understanding of lobbying strategies and dynamics requires time-resolved, system-wide data, which are largely unavailable for most political systems. In the United States (U.S.), the Lobbying Disclosure Act (LDA) of 1995 mandates public reporting of all federal lobbying activities in detailed quarterly filings. However, extracting structured, quantitative information from these filings has remained technically challenging and labor-intensive. Here we present and analyze LobbyView, a relational database that integrates and disambiguates data from more than 1.6 million LDA reports. LobbyView provides access to detailed lobbying disclosures, reconciled corporate entities, and tools for linking LDA data to external legislative and corporate databases. We demonstrate the utility of LobbyView by examining both macro-level and highly granular lobbying dynamics. Specifically, we reconstruct the connectivity patterns of the U.S. lobbying network, and we show how they evolve over time, we identify organizational principles such as the accumulation of professional contacts within a small set of firms, and reveal how lobbying activity is synchronized with electoral cycles. Moreover, we introduce a probabilistic framework for analyzing lobbying behavior at the scale of individual bills, issues, or firms. We envision LobbyView as a resource not only for political scientists, but also for quantitative interdisciplinary research, enabling the application of methods from statistical physics, systems biology, and machine learning to the study of lobbying systems.

Figures (4)

• Figure 1: Our LobbyView database comprises over 1.6 million public records, covering all federal U.S. lobbying since 1999, and enabling a time-resolved reconstruction of the multilayered architecture and scaling behaviors of the lobbying network. (A) The data processing from the Lobbying Disclosure Act (LDA) reports to lobbying network starts with the digitalized text of publicly available disclosures. The atomistic data is then compiled into a relational database, which can be interfaced with other political datasets. To correctly infer the statistics of lobbying activities and infer the lobbying network structure, we clean the data and disambiguate the lobbying actors. (B) Network representation of all U.S. federal lobbying activities during the year 2017. The first layer comprises 10,694 clients initiating lobbying (green), and the second layer comprises 4,405 registrants (brown). The node size in these layers is proportional to the cumulative USD amount spent/received by each individual client/registrant (c.f. Sec. IV and V of mm). The registrants are connected to 11,543 lobbyists they employ (third layer, gray). We also reconstructed historical associations of lobbyists with 128 specific government agencies (turquoise) or current members of Congress (374 legislators; Democrats in blue and Republicans in red). The details of the visualization can be found in mm, Sec. VII.
• Figure 2: The lobbying network is evolving and adapting in response to political landscape and crisis events. (A) Attachment and detachment processes in a sample client-registrant-lobbyist (green-brown-gray) network component during 2020--2023 (for selection methodology see Sec. VIII.D of mm). (B) Number of active K-Street registrants per year (bars) and their yearly increments (circles) and decrements (triangles). The top horizontal bars show the party (Democratic in blue and Republican in red) of the Senate majority, House majority, and President by year. The financial crisis (2007--2008) and COVID-19 pandemic (2019--2021) are shown by gray overlays. (C) Number of active clientships (client-registrant connections) per year (bars) and their yearly increments (circles) and decrements (triangles). (D) K-Street registrant survival probability. (E) Cumulative attachment function $\pi^+(c)$ in terms of K-Street registrant in-degree. The quadratic functional form implies that the probability that a new link 'selects' a registrant with $c$ clients is proportional to $c$. (F) Complementary cumulative distribution (CCDF) of the number of clients per K-Street registrant (K-Street registrant in-degree), with a fitted cut-off power law probability mass function $p(c) \propto c^{-d}\exp(c /c_0)$, where $d = 1.46$ and $c_0 = 86$ (white dashed line). Quantities in (D--F) are computed separately for each year and their precise mathematical definitions can be found in mm, Sec. VIII.
• Figure 3: Our data enables fine-grained analysis of the lobbying network adaptations. (A) Fraction of clients lobbying on a given issue, scaled by its average value in all years considered. Some of the spikes of interest can be easily associated with crisis events, such as the terrorist attacks of 9/11 or the COVID-19 pandemic. (B) Average number of clients lobbying on a given issue. (C) Fraction of clients approaching a given government entity in a given year, scaled by the average value in the period considered. The government entities are listed according to the average number of targeting clients. In this Figure, we present only the most-approached government entities; a more extensive list can be found in mm, Sec. XI. Some of them, such as "President of the United States" (direct lobby) and "Executive Office of the President" (lobby with President's staff), are closely related. Nevertheless, we retain the reported format in order to preserve these subtle distinctions. (D) Average number of clients approaching different government entities. (E) Year-to-year correlation of the issue portfolio vectors (columns of the matrix in panel (A)). (F) Year-to-year correlation of the government approach portfolios presented as columns in panel (C). The correlation matrix has a manifest block-diagonal structure, synchronous with the presidential terms.
• Figure 4: The LobbyView database enables granular analysis of the lobbying dynamics. As an example, in this figure, we analyze lobbying surrounding the Stop Online Piracy Act (SOPA). We restrict our attention to the lobbying strategies of the 67 publicly-traded U.S. firms that lobbied on SOPA in the last quarter of 2011 when the bill was introduced. (A) Technically, our analysis amounts to evaluating the probability of different lobbying pathways. The pathway highlighted in this diagram corresponds to an information sector company, lobbying through both In-House and K-Street firms (mixed profile), which approached the Executive Office of the President (EOP), and whose lobbyists had connections to both of the major parties (bipartisan portfolio). More generally, this diagram can be understood as a coarse-grained version of the lobbying network in Fig. \ref{['fig:network_graph_2017_and_degree_distributions']}B, with nodes corresponding to different compound random variables. The node sizes represent different marginal probabilities, and the edge thickness represents conditional probabilities. (B) SOPA-specific shifts in client sector probability. The sectors correspond to 2-digit NAICS codes. Positive probability shift (most prominent for the information sector, containing IT companies, as well as publishers, broadcasters and motion picture industry) indicates elevated interest in SOPA. (C) IT-specific probability shifts in registrant profile (in-house lobbying department only, both in-house and a K-Street firm (mixed), one K-Street firm, or multiple K-Street firms). (D) IT-specific shifts in the probability of approaching different government entities. Note that multiple government entities can be approached at the same time, so the events in this layer are not mutually exclusive. For brevity, in this figure, we only show the 6 most often mentioned government entities. (E) IT-specific shifts in the probability of having an association with the current legislator of a particular party (Democratic, Republican, both, or none). In panels C-E, we restrict our analysis to the SOPA lobby probability distribution $\mathbb{P}_S$. Full details of the probabilistic analysis are presented in the SI, Section XII.