Organizational context and budget orientations: a computational text analysis

Abstract

Local governments operate in contexts that significantly vary in their complexity, turbulence, and munificence. Such variations in context have important implications for organizational outcomes and practices, including budgetary orientations. To evaluate public sector organizational practices, we focus on budget functions in California county proposed budgets during 2012–2017. These public documents present a wealth of untapped information, which shed light on a number of key organizational variables of interest. Computational text analysis methods offer a highly generalizable means of tapping into public documents in order to generate objective organizational data. Using budget narratives and a general method for analyzing texts offered by the Latent Dirichlet Allocation (LDA) approach, we assess the relevance of organizational context for control, management, and planning budget functions.

Notes

1  As with all text analysis, the fundamental unit of data used in topic models are terms as represented in the document–term matrix. Terms are treated as items from a vocabulary, indexed by a set of numbers {1, …, V}. The vocabulary consists of all the terms in a given corpus, or a collection of documents. A document is a bag of N terms. We describe a document as a “bag of terms” rather than a series or sequence of terms in a particular order because the topic model does not take the order of terms or words into account. These N terms can be represented by a vector w = (w₁, w₂, …, w_N). A corpus is a collection of M documents which can be represented by D = {w₁, w₂, …, w_M}. The topic model treats each document within a corpus as a mixture of a fixed number of k latent topics, each of which is represented by a probability distribution over words.

2 Though not a focus in this study, empirical evaluation of different dimensional reductions might further reveal a more optimal number of topics; for instance, perhaps management content diverges starkly into general management and fiscal management themes, thus, better represented by two distinct management topics. In this way, topic modeling is much like exploratory factor analysis as a means of dimensionality partitioning or reduction.

3 Just as there are many different varieties of factor analysis, there are many different varieties of topic models. The LDA is a basic topic model.

4 The Tausanovitch and Warshaw (2013) political ideology measure is based on Item Response Theory analysis of national survey data, and represents the extent to which each county is politically liberal or conservative on a bipolar axis.

5 These include: dynamic topic models (Blei and Lafferty, 2006; Wang, Blei, and Heckerman, 2012) to analyze the evolution of topics over time and structural topic models which incorporate topic metadata for enhanced interpretability (Roberts et al., 2014).

Additional information

Notes on contributors

L. Jason Anastasopoulos

L. Jason Anastasopoulos ([email protected]) is an Assistant Professor in the Department of Public Administration and Policy, the Department of Political Science and the Institute for Artificial Intelligence at the University of Georgia. His research focuses on representative bureaucracy and political methodology. His work on political methodology focuses on modern causal inference, text analysis and image analysis and his work on bureaucracy focuses on race, ethics and decision-making in representative bureaucracy. Current topics of interest include fairness and ethics in machine learning, artificial intelligence and governance, the behavioral micro-foundations of the theory of representative bureaucracy and causal inference methods for high dimensional data.

Tima T. Moldogaziev

Tima T. Moldogaziev ([email protected]) is an associate professor in the School of Public Policy at Pennsylvania State University. His primary research interests are in public sector management, regional and local governance, public sector infrastructure financing and fiscal policy.

Tyler Scott

Tyler A. Scott ([email protected]) is an assistant professor in the Department of Environmental Science and Policy at University of California, Davis. His primary research interests include collaborative governance, infrastructure and public service delivery, policy networks, and the role of science in policy.