New Way to Author Questionnaires for the U.S. Census Bureau
The Business – Creating a Snapshot of the U.S. Economy
Every five years, the Economic Directorate of the Census Bureau conducts an Economic Census, creating a profile snapshot of the U.S. economy. The Economic Census gathers data from all large- and medium-sized firms, and from a sample of small businesses.
Unlike the better-known Decennial (population) Census, which involves only a handful of different questionnaires, the Economic Census requires more than 600 questionnaires in order to handle the broad diversity of the U.S. economy, including manufacturing, mining, construction, wholesale, retail, and other services.
The Challenge – No Way to Share Metadata across Questionnaires
Because of the dynamic nature of the economy, the Census Bureau needs to redesign the many Economic Census questionnaires every Economic Census. Some categories of business grow and require more detailed information (e.g., microbreweries), and sometimes new categories emerge that never existed before (e.g., cloud computing service providers.)
Although the many questionnaires are distinct, questionnaires share many aspects. For example, the RT-44201 questionnaire (Furniture Stores) shares a great deal of metadata (e.g., instructions, other text, and response data definitions) with the RT-44202 questionnaire (Home Furnishings Stores), as one might imagine. Yet the existing questionnaire design applications had no way to combine these shared metadata (data about data) in any way.
The Solution – Reusing Metadata & Automated Generation of Pages
Fenestra Technologies developed a solution that would leverage the use of metadata to allow for reuse among questionnaires as well as generating both conventional paper and electronic versions of the questionnaires. Another goal of the system was to allow fully automated generation of questionnaire pages wherever possible.
Fenestra’s solution is known as the Generalized Instrument Design System (GIDS). GIDS comprises six modules:
- An autoformatter: The autoformatter automatically generates rule-based page layouts from metadata without manual intervention.
- A custom layout designer: This module draws metadata from a central repository and allows a designer to specify the location and formatting of text response areas within a question.
- An assembler: The assembler combines both manual and automated question layouts into complete questionnaires with automatic pagination and header and footer handling.
- A previewer: The previewer provides WYSIWYG on-the-fly previews of both individual question layouts and completed questionnaires.
- A publisher: The publisher generates print-ready questionnaire files (PostScript or PDF), as well as form definition packages used by the Surveyor electronic reporting application.
- A behavior editor: In addition to mimicking the general layout of paper questionnaires, electronic questionnaires also offer dynamic data generation and validation capabilities, such as automated computation of sums of values and sanity checks. We refer to these functions collectively as “behaviors,” and the behavior editor allows the questionnaire designer to tailor the behaviors as required by the different questionnaires. (Typically, an Economic Census electronic questionnaire includes nearly 400 behaviors, roughly 80% of which are generated and managed automatically.)
The Results – Widely Used Digitized Questionnaires & Millions in Savings
The use of Surveyor-GIDS electronic reporting contributed to an overall processing savings of $1.6 million for the 2007 Economic Census. On average, metadata fully generated automatically about half of the page layouts in the Economic Census. Of the remaining 610 manually created layouts used during the 2007 Economic Census, multiple questionnaires reused 257 layouts, with a mean level of reuse of 21 questionnaires per layout.
Furthermore, the success (and generality) of GIDS is exemplified by its increasingly widespread applicability: It is now also being used to create questionnaires for several other surveys, with more on the way, and even for preparing and cataloging other kinds of non-questionnaire documents (e.g., letters and information sheets.)
Mission Critical Data Collection Server for the U.S. Census Bureau
Fenestra Delivers on Mission Critical Project
The Economic Directorate of the Census Bureau is responsible for the collection and publication of certain business survey data. The Economic Directorate collects data through surveys and censuses, including an Economic Census, which the Economic Directorate conducts every five years. As part of an effort to reduce respondent burden, increase data quality, and reduce data collection cost, the Economic Directorate and Fenestra initiated a project called the Generalized Instrument Design System (GIDS) in 2000. As part of this project, Application Services Division (ASD) of the Census Bureau developed and maintained a data collection server. The Census Bureau utilized the server from the inauguration of the GIDS Program in 2001 through 2011.
In early March 2011, the Census Bureau decided to transfer responsibility for the data collection server to the Electronic Reporting Branch (ERB) of the Economic Programming Division (EPD). ERB was unable to use the existing code base and needed to create a new data collection server system for the 2011 Annual Survey of Manufacturers and Report of Organization (ASM/COS).
The data collection server is a mission critical portion of the electronic reporting work flow. Due to workload and time constraints, ERB required a fully functional replacement server immediately so that the Census Bureau could fully test the new data collection server system prior to mail-out of the 2011 ASM/COS in December 2011.
ERB researched new requirements for the system and found that the technologies involved were orthogonal to ERB’s existing domain knowledge. Schedule and resource constraints impeded EPD being able to implement the necessary technologies in the time frame required. The risks associated with a delay in completing a robust server on time were potential schedule slippage and data loss, both critical matters in Census Bureau data collection.
ERB consulted with Fenestra over the course of a month while attempting to reverse engineer the existing system. After several consultations, Fenestra offered to investigate and undertake the process of reverse engineering the primary data collection server and providing a non-complex method for ERB staff to tie in the data collection server to the Census Oracle data repository.
Fenestra Delivers Timely Results
ERB accepted the offer, and Fenestra began work. In spite of other project commitments, Fenestra was able to design, implement, test, and deliver to the Census Bureau a ready-to-use data collection server in less than one week.
This rapid turn-around was achieved because of the diverse domain knowledge and extensive development experience of Fenestra’s staff.
ERB implemented the connections to the Census Oracle data repository, and the Census Bureau fully tested the system for use in the 2011 ASM/COS without any further modifications required by Fenestra. The Census Bureau eliminated a potential schedule risk and verified the reliability of the data transfer mechanism. Fenestra delivered a solution to the Census Bureau at minimal cost without compromising Fenestra’s other commitments to the Census Bureau.