State and Local Government Statistics at a Crossroads (2007)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

5 Dissemination and Analysis T he ultimate objective of the myriad tasks that the Census Bureau’s Governments Division undertakes in its state and local government statistics program is to provide useful and accurate data in a timely and readily accessible format to data users. In this regard, the division receives mixed reviews. While its data are widely viewed as relevant and accurate, the long delays in dissemination of several of the data series re- duce their usefulness to the user community. Moreover, the knowledgeable Governments Division staff carries out very little data analysis, which could be helpful to users. The division has made great strides in improving elec- tronic access to the data, and much more in the way of enhancing access is on the drawing board. This chapter addresses the three areas of timeliness, website access, and in-house analysis for the state and local government data series. TIMELINESS All types of users with which the panel interacted—federal government agencies, nonprofit organizations representing state and local government interests, and academic researchers—said their greatest concern was the lack of timeliness of the data, and the panel agrees with this concern. The fact that information on state and local government finances is not available until the third annual revision to estimates of gross domestic product (GDP) that were initially released two years earlier is unacceptable, considering 94 

DISSEMINATION AND ANALYSIS 95 the significant percentage of GDP that the state and local government sec- tor contributes. It has been suggested that the voluntary nature of the state and local government surveys may contribute to the lack of timeliness. Other factors that may play a role in delaying data releases include the complexity of the information collected, some of which requires special gathering and tabu- lation; the difficulty of communicating with some of the small local and special governmental bodies that are asked to complete the questionnaires; data collection procedures, especially the joint Governments Division–state arrangements for central collection (see Chapter 4); and procedures to pro- cess the data once they are received. Timeliness varies across the division’s surveys (see Table 5-1). For example, reports from the quarterly tax and retirement system surveys are fairly timely, but not so timely as quarterly reports from other divisions in the Economic Directorate, such as the Quarterly Financial Report (Manu- facturing) and the Quarterly Service Survey: the latter two reports come out 75 days after the end of the quarter, while the tax and retirement system survey reports come out 90 days and 120–150 days after the end of the quarter, respectively. The Governments Division has considered revising the structure of the quarterly tax survey to make it possible to release the data more quickly, in order to capture an “instant view” of what is happening to state and local government tax revenues. One approach would be to convert the current quarterly collection program into a monthly collection with less detail, taking into consideration the concern for timeliness as well as accuracy and completeness. The annual state and local employment survey data are released about 12 to 13 months after the March reference date, which the Governments Division argues is a fairly reasonable time frame. A principal reason for the delay in getting the data processed and disseminated is the difficulty in obtaining responses to the survey—whether in paper or electronic for- mat—in a timely manner. The division has investigated ways to expedite the reporting of data. One interesting idea is to coordinate with major software vendors (e.g., Peoplesoft) to build the survey into the enterprise information systems that they install in state and local governments. This effort has not met with much success to date. The survey with the greatest timeliness problem is the annual finance survey. The finance survey has several different data releases—for state and local public employee retirement systems, state finances, public elementary and secondary education finances, and state and local government finances. The state and local government finance survey, which typically is released by the Governments Division 21 months or more after the June ending   As stated by Henry Wulf during his presentation at the panel’s June 2006 workshop. 

96 TABLE 5-1  Timeliness of Release of Governments Division Surveys, as of April 2007 Survey Reference Date Typical Months to Release Notes Quarterly Survey of State 3 calendar months ending Within 3 months of the and Local Government March, June, September, and reference quarter ending Taxation December. Quarterly Survey of Public 3 calendar months ending Within 4.5 months of the Employee Retirement System March, June, September, and reference quarter ending Finances December Annual Survey of Pay period including March Within 12 months of the Government Employment 12 reference pay period Annual Survey of State State government fiscal years Within 9 months of the Government Tax Collection ending June and four states reference fiscal year ending closest to June: New York (March), Texas (August), and Alabama and Michigan (September) 

Annual Survey of Public State and local government Within 15 months of the Data could span nearly 2 years. Employee Retirement System fiscal years ending between reference fiscal year ending For example, totals for fiscal years Finances July 1 and June 30 (e.g., July ending between July 1, 2005, to June 1, 2005, to June 30, 2006). 30, 2006, include, at the extremes, Exceptions include some governments with fiscal periods of governments outside this August 1, 2004, to July 30, 2005, reference period (e.g., Texas, and governments with fiscal periods Alabama, and Michigan of July 1, 2005, to June 30, 2006. state governments), which In effect, the data reflect economic are included with the conditions toward the latter half of closest reference period the reference period, because about (e.g., September 30, 2006, 40 percent of governments have fiscal Michigan state data included years that end in December and 40 with July 1, 2005, to June percent have fiscal years that end in 30, 2006, governments data) June. Annual Survey of State Same as Annual Survey of Within 18 months of the Data could span nearly 2 years as Government Finances State Government Taxation reference fiscal year ending above. Annual Survey of State and Same as Annual Survey of Within 23 months of the Data could span nearly 2 years as Local Government Finances Public Employee Retirement reference fiscal year ending above. System Finances SOURCE: U.S. Census Bureau, Governments Division staff. 97 

98 STATE AND LOCAL GOVERNMENT STATISTICS AT A CROSSROADS period, is the slowest of the finance surveys. This lack of timeliness has a negative impact on the national accounts. In producing the national income and product accounts, the Bureau of Economic Analysis (BEA) prepares annual estimates based on the Census of Governments and the annual fi- nance survey. Because of the data lag, the data are fully applicable only for the third annual revision of GDP. In producing the second and first annual revisions, BEA uses other data that are available at the time of the produc- tion of the estimate and extrapolates from other components. The quarterly estimates are based on some quarterly data (compensation, construction, certain social benefits, and tax receipts) and are otherwise interpolated from annual estimates. A complicating factor that exacerbates the timing problem is the varia- tion in fiscal years between states and localities. For example, the 2004 annual finance data were posted to the Governments Division website in August 2006, 26 months after the close of the typical state government fis- cal year ending June 2004. However, for many local governments, the time lag was longer. The 2004 annual finance survey includes fiscal years ending between July 2003 and June 2004, which means that the survey includes FY 2003 data for many local governments. For a local government with a calendar fiscal year, the data were released 32 months after the December 2003 fiscal year end. For a local government with a fiscal year ending in September 2003, the time lag was 35 months., The Governments Division has taken steps to speed the processing of the estimates once they are received in the Census Bureau. By late 2005, the division fully implemented E-Basic, a 4-year project to reengineer how the division compiles, edits, and adjusts finance data for all state governments, and began expanding it to the largest local governments. Among other ac- complishments, E-Basic developed a highly flexible computer application   Many users undoubtedly are not aware that the 2004 finance data actually are for the 2003 fiscal year for many local governments. As a result, they may not understand that the lag is as long as it is, and probably more important, they may believe they are analyzing different time periods than is the case. The Governments Division discloses the time periods covered in its documentation, although perhaps not as prominently as needed.   Audited financial statements and associated Comprehensive Annual Financial Reports (CAFRs)—standardized reports that are used by rating agencies and bond markets to deter- mine the solvency of the governmental unit—are based on detailed data that often overlap with the detailed data governments provide to the Governments Division in support of the annual finance survey. CAFRs usually are available sooner than Governments Division data, but not always dramatically so. While CAFRs often are expected or required to be released by state and local governments no later than 90 days after the close of a fiscal year, and many governments achieve this, even large governments sometimes take 9 months or more to release their CAFRs. Smaller local governments sometimes take much longer than this to release their CAFRs. Furthermore, the supporting details needed by the Governments Division are not always available in usable form until after CAFRs are released. 

DISSEMINATION AND ANALYSIS 99 that can process the idiosyncratic and detailed record-keeping systems of different state governments, enabling division analysts to cease focusing on detailed transactions and instead concentrate on the overall accounting systems. This has meant that in analyzing expenditures for a typical state in a given year, an analyst would have to code only 400 new items instead of 30,000. E-Basic has reduced the completion time for processing individual state data by 75 percent, from approximately four weeks to one. The proj- ect team won a Census Bureau director’s innovation award for developing and implementing E-Basic. With this new system in place, the key obstacle to timeliness now appears to be the slowness of data availability from state and local governments. E-Basic has speeded up data processing for the core of state govern- ment, but other elements of the process can delay processing and release of data. A government, as defined by the Governments Division, usually is much more than the core government that typical users think about. Large governments usually contain several or many component units with inde- pendent accounting and financial reporting systems, such as universities, transportation authorities, and other entities, and they are not included in the E-Basic system. Even when data from the core entity are received and processed quickly by the Governments Division, it can be many months or more before data for the component units are obtained and processed. The Governments Division has several options under consideration to deal with the lags in availability of the finance survey data. The division has considered releasing data as they are processed, thus changing the process- ing system to release groups of 10 or 15 states. This approach would still delay the production of a final national total until data from all govern- ments were available or unless the national total could be provisionally estimated and released as a preliminary estimate. Another approach would be to make greater use of preliminary estimates, so that estimates for the nation and all states could be released more quickly, with some of the esti- mates consisting of provisional figures. Preliminary releases might provide opportunities for users to help the Governments Division find errors and anomalies in the data before the final data are disseminated. In addition to staggered releases of partial data and preliminary estimates, the division has considered increased use of unaudited data. Although the division cur- rently accepts unaudited data, state governments are reluctant to provide unaudited data because they do not want two sets of data released to the public. To guide decisions on methods of releasing preliminary or partial data from the annual finance survey, the division will need to conduct research on the costs and benefits of alternative estimation and dissemination meth- ods. In the case of preliminary estimates, such research will require mod- eling with prior-year data to simulate the effects of preliminary estimates 

100 STATE AND LOCAL GOVERNMENT STATISTICS AT A CROSSROADS developed using different estimation methods. As part of this work, the division should investigate the methods used by other statistical agencies that release preliminary or partial data in response to user needs and how those agencies inform users of the properties and appropriate use of pre- liminary and final estimates. A small representative national sample could also be considered. Recommendation 5-1: The Governments Division should give high priority to a program of research on the benefits and costs of adopt- ing earlier release procedures for the annual finance survey and other surveys by such methods as releasing preliminary estimates or releasing estimates as they are compiled. The research should include evaluation of the ability of preliminary releases to replicate prior-year data and analysis of preliminary-to-final differences attained by using different estimation techniques. GOVERNMENTS DIVISION WEBSITE The Governments Division has made a major commitment to using the Internet as the primary means of disseminating its data to the public, but developing a useful website for users with different needs and levels of experience and expertise is a challenging task. To gain a general sense of useful strategies for addressing the Internet access needs of researchers who require detailed data, the panel heard from the Weldon Cooper Center for Public Services at the University of Virginia, which has worked on ways to make the data from various federal agencies available in easy-to-use formats on the Internet. Considerations in Facilitating Researcher Access The Cooper Center notes that researchers who use federal data face the challenge that there is no standard way of storing or transferring data. Different agencies store data and metadata in different ways, making it dif- ficult for them to release data in standard ways that facilitate transferring the data into statistical programs for analysis purposes. (Metadata—the contextual background information that is necessary to interpret a data- set—inform users of who created the dataset, when it was created, how the underlying data were generated, the statistical assumptions built into the sample design and estimation for the data, and similar information.) The Census Bureau publishes data in portable document format (PDF), which   Discussion of the work of the Cooper Center is based on William Shobe’s presentation at the panel’s June 2006 workshop. 

DISSEMINATION AND ANALYSIS 101 makes it hard to transfer the data into statistical programs and to map to metadata sources. In the view of the Cooper Center, agencies should avoid assuming that they know how their data will be used. There are researchers who may not know what data they want and need until they know what is available, which makes the ability to conduct creative searches of data and metadata important. Toward this end, the Governments Division should provide the widest array of data sources and focus on distribution rather than layout of its data. It is important for researchers to be able to extract the data they need as efficiently as possible, without language barriers between the database and their computer system or dependencies on one operating sys- tem. Automatic feeds, such as really simple syndication (RSS), a popular online technology that publishers use to distribute their content to other sites and services, could allow researchers to subscribe to periodic data and automatically update their models as soon as data are released. This type of arrangement would lower transaction costs for researchers. Currently, the Governments Division data are generally found behind a “20-click labyrinth,” meaning that researchers must spend considerable time selecting the series, data items, and format they want each time the data are updated. The metadata are often not attached to the data, requir- ing researchers to copy and paste text from another webpage. Although Excel and comma delimited files are easier to manipulate than PDF files, they still require human intervention. A direct process that allows research- ers to download the data directly into their analytical software would not require human intervention. Improving the Governments Division Website The Governments Division has begun to address data dissemination in terms of website development as part of a broader Census Bureau ef- fort. The Census Bureau has established a Web Council with an executive guidance group, in order to facilitate making data more accessible in a consistent way across all programs. Part of the greater Census Bureau plan is to allow users to access the data from individual programs more easily without needing to understand the bureau’s organizational structure. The Census Bureau’s outreach in this area has been extensive—for example, it is working with and learning from the Australian government as it develops a national data network. The Governments Division is one of the early programs going through the process of coordinated website development at the Census Bureau. Given its early start, the division has the opportunity to be a part of setting the standards in this area for the Census Bureau as a whole. The division sees three groups of users who are asking for data from 

102 STATE AND LOCAL GOVERNMENT STATISTICS AT A CROSSROADS the Internet—the sophisticates who know exactly what they need, the neophytes who are sophisticates in other fields but do not understand state and local governments, and the occasional users who have a general under- standing from having used the data before and need to ask questions and obtain specific information. The website development is striving to meet four criteria—simple, streamlined, easy, and accurate—to satisfy these three user groups. Metadata are to accompany all information. The proposed redesign for the Governments Division Internet site is an adaptation of a Bureau of Justice Statistics Internet site (http://www.ojp. usdoj.gov/bjs/) that incorporates many features that users are requesting. For data access, the Governments Division has selected an extraction sys- tem to provide three dimensions easily—geographic (national, state areas, individual governments), temporal (the most recent available year or data for several years), and data detail (totals or components of aggregates). High-end users can download the detail. Medium- and low-end users will see a table shell with sufficient labeling and warnings that they can populate with data they can manipulate for analysis. The division is in the process of evaluating whether a table function or a build-a-data-set function will be more useful; it has not made decisions on what types of files will be downloadable (e.g., SPSS or SAS). The build-a-data-set capabilities have been employed to the benefit of researchers by both the Bureau of Labor Statistics and the Bureau of Economic Analysis, which now provide the capability of generating panel data on their websites. As a government agency, the Governments Division faces several stric- tures on the development of its website. The division needs to follow Census Bureau rules concerning the statistical validity of data, and the downloading and table generating components of the system must provide metadata whenever appropriate. The division also must follow Census Bureau rules concerning the look and feel of the website. Finally, the site must be Section 508 compliant, that is, it must provide data accessibility for people with disabilities. The website improvement project is being conducted in three phases: • Phase I: In this first phase, the Governments Division continued to redesign the look, feel, usability, and navigation capabilities of its website. A major milestone was the introduction of metadata to accompany releases of data. The division also plans to introduce new user tools for data dissemination on the site, such as Build-a- Table and Build-a-Data Set. This phase is likely to be completed in 2007. • Phase II: The Governments Division plans to introduce new search capabilities and new methods to enhance navigating through the website. The end of this phase will see the introduction of new 

DISSEMINATION AND ANALYSIS 103 user tools, such as graphing and data comparison capabilities. This phase is likely to be completed several months after Phase I is completed. • Phase III: Work in the final phase will consist of putting complete production capabilities on the website that are designed to ease navigation, increase usability, and make metadata more accessible. Any remaining access tools will be added at the end of this phase, which is expected to be completed about six months after Phase II. Recommendations The panel fully supports the project to improve the Governments Divi- sion website so as to facilitate access to its data and associated metadata. Users now expect data to be available on the Internet, and a well-designed website can facilitate access, thereby building a user community and the support for an agency’s data programs. The panel is pleased that attention is being paid to how to provide access on the Governments Division website, not only to files of detailed data for individual governments, but also to table-generating capabilities for users who need selected information in table format. The panel urges the division to think creatively about the design and linkage capabilities of the site. For example, the site should permit ready links to other Census Bureau websites that provide detailed population and private-sector busi- ness data for states, counties, cities, towns, townships, and school districts. The site should also include mapping capabilities, graphical analysis tools, and tabulation systems to facilitate comparisons over time and among governmental jurisdictions. The panel understands the requirements for the Governments Division website to conform to Census Bureau standards and protocols. However, the panel thinks that some flexibility in website design for specific units, such as the Governments Division, should be encouraged to accommodate special features and data needs that are particular to the unit. In addition, the Governments Division website should be made easier to find from the Census Bureau’s home page and other pages than at present. At the mo- ment, the governments data are lumped in with economic and industry data, presumably because the Governments Division is located in the Eco- nomic Directorate. Yet the user communities for the data on state and local governments are different from those on private-sector businesses, so that every effort should be made to clarify the access paths to the state and local government data. As the Governments Division completes the shift away from paper-based dissemination to web-based access, care must be taken 

104 STATE AND LOCAL GOVERNMENT STATISTICS AT A CROSSROADS to ensure that the data are complete and available. Finally, there must be provision for user feedback about the pluses and minuses of the website and for continuing improvements in the design and content of the site and in data access tools and links to other relevant data. Recommendation 5-2: The Governments Division should continue to give high priority to the redesign and continuous improvement of its website. There should be clear access to the site from the Census Bureau’s home page and other access points. Desirable features of the site include: • metadata (information about the data) that are complete and easy to access, including use of hot links to information on definitions and measures of error of the type that has traditionally appeared in Governments Division publications; • the capability to crosslink and combine state and local government data with data from other sources, initially with data from other Census Bureau data series on population and industry for states and local areas and, in the future, when issues of data comparabil- ity are resolved, with data from other federal agencies; and • graphical analysis and mapping tools to facilitate comparisons over time and among jurisdictions. IN-HOUSE ANALYSES Despite the fact that the Governments Division staff have a wealth of expertise on such topics as state and local government organization, finance trends, and employment trends, and that the staff possess a depth of analytical skills that are honed in the difficult tasks of data editing and imputation, the division does little analysis of its own data for public con- sumption. Like its sister organizations in the Economic Directorate, the division usually releases Census of Governments and survey results without descriptive analysis or graphics. For the Governments Division, this practice dates back to the early 1990s, when most descriptive and analytical reports were dropped and the division began placing data on the Internet with little or no explanatory text and giving readers relatively little ability to compare data among governments or over time. In fact, the division has increasingly turned to web-based dissemina- tion. The 2002 Census of Governments will be the last printed version.   Currently, state and local government finance data are not available online for years prior to 1991-1992. 

DISSEMINATION AND ANALYSIS 105 The release of data to the public has become even more of a low-key op- eration, with new releases of data announced in limited distribution news alerts that inform users of the availability of new information rather than press releases that announce the new information and provide analytical highlights. This practice of releasing data to the public devoid of fanfare, major findings, analytical commentary, or key contacts on the staff means that the Governments Division and the Economic Directorate generally leaves it to data users to cull through the division’s website and to draw inferences and meaning from the flood of state and local government data that are issued on an annual and quarterly basis. As a result, although many data users obtain their state and local government information directly from the Governments Division and develop their own analyses, many others are turning to a cottage industry that has emerged in the private sector and among a few public interest groups for information that is tailored to their needs and interests. The panel learned in discussions with major data users that much of the needed value-added information from the Governments Division could be quite simple and straightforward to provide and could even be largely automated. Users generally want simple derived measures, such as comparisons over time, time series presented in inflation-adjusted dollars, per capita amounts, and amounts in relation to personal income that the Census Bureau could comfortably provide. Other derived measures, such as rankings, are less obviously candidates for development by the Govern- ments Division, since they are often the subject of political rather than analytical interest. It is worth noting that Statistics Canada press releases provide these types of value-added information to users and offer an ex- ample that the Governments Division should consider. In addition, the division staff could prepare accompanying text regarding trends over time and among governments—such as trends in different regions of the coun- try—that would provide context for users. Occasionally, the staff could prepare longer analytical pieces, similar to those that are prepared by the Bureau of Labor Statistics staff in the Monthly Labor Review and by BEA staff in the Survey of Current Business.   The release of the long-awaited state and local government finances data for 2003–2004 is a case in point. The data were announced with only a Census Bureau publication called a “Tip Sheet” and a reference to the downloadable Internet tables in an advisory on May 31, 2005.   Statistics Canada press releases are located on its website at http://www.statcan. ca/menu-en.htm. 

106 STATE AND LOCAL GOVERNMENT STATISTICS AT A CROSSROADS Recommendation 5-3: The Governments Division should add value to the data that are released on its website by providing simple derived measures, such as per capita expenditures and taxes, more explanatory material, and comparative contextual analyses—for example, of trends by type of government and region. The division should also facilitate wider dissemination of its data by regularly issuing press releases that include statistical comparisons with previous data.