TreeCovery: Coordinated dual treemap visualization for exploring the Recovery Act

Abstract

The American Recovery and Reinvestment Act dedicated $787 billion to stimulate the U.S. economy and mandated the release of the data describing the exact distribution of that money. The dataset is a large and complex one; one of its distinguishing features is its bi-hierarchical structure, arising from the distribution of money through agencies to specific projects and the natural aggregation of awards based on location. To offer a comprehensive overview of the data, a visualization must incorporate both these hierarchies. We present TreeCovery, a tool that accomplishes this through the use of two coordinated treemaps. The tool includes a number of innovative features, including coordinated zooming and filtering and a proportional highlighting technique across the two trees. TreeCovery was designed to facilitate data exploration, and initial user studies suggest that it will be helpful in insight generation. RATB (Recovery Accountability and Transparency Board) has tested TreeCovery and is considering including the concept in their visual analytics.

Introduction

In February 2009, President Obama signed an economic stimulus package into law, dedicating $787 billion to create jobs and boost the economy, with the provision that the distribution of the money would be completely transparent. The growing recognition of the importance of design excellence in e-government applications (Fedorowicz & Dias, 2010) has raised attention to models for measuring user satisfaction (Verdegem & Verleye, 2009) and usability guidelines for e-government websites (Donker-Kuijer, de Jong, & Lentz, 2010). These concerns have taken on increased importance as the Obama administration expands its data availability efforts under the Open Government Directive. In fulfillment of this requirement, the agencies in charge of distributing the money and all recipients issued periodic reports detailing how the money they controlled was spent. These publicly available reports comprise a large amount of data, containing information about the effectiveness of the stimulus package, the general trends of distribution, and potentially interesting outliers.

Some effort has already been expended toward producing visualizations of this data that could assist in revealing such details. The government commissioned a website, Recovery.gov, dedicated to this purpose, and several independent journalism outlets have produced their own applications, all offering a particular take on the data. Most of the existing visualizations consist primarily of either tabular or geographical displays. The goals of this effort were to geographically display the distribution, allocation, and expenditure of stimulus recovery funds.

While the data lends itself well to geographical layout, given that states and counties are convenient schemas for chunking the data, exclusive use of maps cannot adequately portray alternate views of the monetary distribution. Specifically, money was distributed through 28 agencies, which assigned it to projects at their discretion; funding was placed in the charge of the prime recipient, who in turn funded sub-recipients and/or vendors as necessary for the project. Agencies naturally funded projects nationwide, and recipients for each project were not necessarily all located in the same area. This view of the data—an agency > project > recipient hierarchy—cannot be adequately conveyed by a geographical substrate.

Our tool, TreeCovery, offers a way to explore data both geographically and according to the monetary outlays. TreeCovery accomplishes this goal through the use of two coordinated treemaps, one drawn with a geographic hierarchy and the other one with levels corresponding to the agency > project > recipient money flow. While the views presented by the two treemaps differ, the underlying data remains identical at all times. Filtering is coordinated across the views and a proportional highlighting technique is used for coordination.

In addition to the coordinated treemap design, we incorporated a few other features to improve exploration techniques. From news articles about the Recovery Act we found many of them using demographical statistics such as population or unemployment rates. We thus included census data for each county, and made it possible to filter by demographic attributes. We also added the ability to save snapshots of the current state of the treemap for3 later comparison. Finally, we included support for emphasizing invalid data values.

The goals of this effort were to enable users to

• Explore the allocation of stimulus recovery funds by agency and states/counties;

• Identify extreme cases such as concentration of spending by an agency in one state or county;

• Discover unusual patterns of spending that show inequities by region, state, and county;

• Understand which agencies were most active in their state or county;

• Facilitate error detection and omissions in the data.

Of course, a powerful interface would also enable other tasks, which were beyond the initial planning of the designers. The larger goals are to empower policy makers, journalists, and citizen groups to have increased capacity to explore key data sets that are tied to national priorities.

TreeCovery has been tested by RATB and received positive feedback about the concept. Hopefully features of TreeCovery will be included into next version of RATB's visual analytic platform.

Treemaps are among the growing set of information visualization tools that could increase the analytic capabilities of government agency staffers, political analysts, journalists, and other interested citizens. The capacity to identify interesting patterns, clusters, gap, outliers, and others features is increasingly important in detecting fraud, ensuring fair allocation of resources, and refining policies to ensure effective use of public funds.

Section 2 discusses related work, while Section 3 provides an explanation of our analysis process, including a detailed illustration of Spotfire's² ability to support exploration of the Recovery Act data. Section 4 explains TreeCovery in detail, while Section 5 offers some sample insights found by the tool. 6 Future work, 7 Implications and recommendations suggest future work and offer conclusions.