The federal statistical system is in alarming disarray. Unless Congress implements broad reforms, the accuracy of the nation's statistical data-so vitally important to sound policymaking as well as to the general functions of government, businesses, and other institutions-is in danger of being compromised by the system's inability to keep pace with America's rapidly changing society.

In an economy that increasingly depends on and is driven by information, flawed data can impose enormous costs, leading businesses and agencies to miscalculate and make the wrong decisions and policymakers to advance ill-advised legislation with disastrous long-term consequences. Moreover, the impact on the federal budget could vastly exceed the budgetary costs of all the current federal statistical agencies combined. Agencies like the U.S. General Accounting Office (GAO) over the past few years have issued reports that detail the system's problems and outline recommended reforms, but little has been done. Consequently, many Americans have begun to question the quality of the data being collected and how those data are being used to prepare the statistics produced by the U.S. government.

At the root of the problem lies an extremely fragmented structure of overlapping agencies and purposes that is as inefficient as it is ineffective. This system generally disseminates two types of statistics: public statistics, which are of interest to the public and business community generally, and administrative statistics used primarily within the government for policymaking and rulemaking. Producing and publishing these statistics on the country's economic and social makeup involves 70 different agencies within 12 Cabinet departments (see Table 1). According to the Office of Management and Budget (OMB), the entire federal government is spending close to $2.7 billion each year on these statistical operations.¹

The structure of the U.S. statistical system is one of the most decentralized in the world, over and above the hundreds of private, state, and local producers of statistics like Dun & Bradstreet and F. W. Dodge Inc. The work of this fragmented federal system is loosely coordinated by a small group of people within OMB known as the Statistical Policy Branch (SPB).²The SPB has the ability to set classification and quality standards; its ability to assert its leadership and require action by individual agencies, however, is limited by the highly decentralized nature of the overall system. In other words, the system is being undermined by its own decentralized structure, despite oversight by the SPB and despite sizable funding. This structure hinders critical improvements and permits agencies to spend resources on duplicative bureaucratic overhead. As a result, the quality of the nation's data deteriorates, and further erosion cannot help but have a damaging effect on the federal budget, a detrimental effect on the public policy debate, and disastrous implications for businesses.

Members of Congress who already recognize these problems have begun to examine ways to consolidate and de-politicize the highly fragmented, costly, and duplicative statistical system. In early 1997, the Senate Governmental Affairs Committee and House Government Reform and Oversight Committee conducted hearings on the system and proposals to consolidate it. While consolidation alone may not be enough to solve the problems plaguing the federal statistical system, it is equally true that these problems cannot be addressed effectively as long as the current structure remains in place. Legislation to restructure the system may be introduced in the Senate and the House later this year.

As a first step in consolidating the many agencies conducting the nation's statistical activities, Congress should consider folding the Bureau of Labor Statistics (BLS), Bureau of Economic Analysis (BEA), Bureau of the Census, and National Center for Health Statistics (NCHS) into a new Bureau of National Statistics (BNS). The new agency should be headed by an independent chief statistician, and a board of directors should be appointed to oversee its activities. As the nucleus of a new federal statistical system, it should be given primary responsibility for the production and accuracy of the nation's public and administrative statistics. The many remaining agencies should be incorporated into the BNS over a period of seven to 10 years.

Consolidating all the statistical agencies into a Bureau of National Statistics would have many advantages. It would free wasteful and duplicative resources. It would ensure a greater level of objectivity, reliability, and insulation from possible political manipulation. It would allow the federal government to develop-for the first time-a coherent national strategy on producing and publishing the nation's statistics. And it would improve the quality of the nation's data while protecting the privacy of respondents.

Regardless of how Congress chooses to reform the federal statistical system, it also will be extremely important to maintain effective oversight both in Congress and in the White House Office of Management and Budget. OMB should comment on classification and quality standards, continue to solicit public comments, and review and approve all federal information collection. Congress also needs to ensure that the goals and objectives of public laws are met, including efforts to minimize the paperwork burden and provide more useful information to the public.

A SYSTEM IN SHAMBLES

The federal government's use of statistical data is not new. The framers of the U.S. Constitution mandated that the government conduct a population census once every decade as the basis for reapportionment of the House of Representatives. Over time, new needs for statistical data have surfaced, and Congress invariably has responded by creating separate statistical agencies throughout the government.³In addition, the new information age has ushered in successive advances in computer technology and access to information worldwide through the Internet. These new capabilities portend that the role of statistical data in public policy and the economy will become more important than ever. Failing to implement necessary improvements in the system therefore will lead to distortions in the data provided to the public and private sectors, with costly results. Good statistics are critical, for example, in determining the most efficient location for new roads, airports, or manufacturing plants, and in deciding when to build them. They also are vital in planning successful product development and marketing strategies. Bad data, on the other hand, can cost billions of dollars each year in misspent resources and lost job opportunities.

Various government watchdog groups have acknowledged both the problems within the federal statistical system and the profound impact of faulty data on the federal budget and policy debate. In July 1995, the General Accounting Office published a report entitled Economic Statistics: Measurement Problems Can Affect the Budget and Economic Policymaking. In December 1996, the Advisory Commission to Study the Consumer Price Index reported to the Senate Finance Committee that just one source of upward bias in the Consumer Price Index (CPI) could have added over $271 billion to the national debt between 1975 and 1996.⁴A faulty CPI affects the accuracy of statistics on real economic growth, productivity, poverty rates, and real wages. This in turn affects the public policy debate by, for instance, misleading Members of Congress and the President into thinking that the actual growth in these statistics is slower than it really is. Moreover, distortions in the measure of the number of families in poverty affect both specific program budgets and public policy in general.

Particularly disturbing is the declining quality of the decennial census. Despite the vital constitutional role of the census and its use in allocating federal funds for education, welfare, and transportation, the GAO admitted in 1992 that the 1990 census was inaccurate.⁵For the first time since the inception of the national census, the quality of the data produced had deteriorated in relation to a previous census. This deterioration occurred despite the fact that the 1990 census was the most expensive in history, costing $2.6 billion. Moreover, the outlook for improvement is bleak: A recent GAO study concluded that the 2000 decennial census is a high-risk project that is likely to produce unsatisfactory data.⁶

Two of the most important consequences of the fragmented federal statistical system are the deterioration of economic and social statistics and the subsequent decline of public trust in federal survey results. These problems can be attributed to fragmented structure, a lack of accountability, the absence of a national statistical strategy, and a highly politicized statistics-gathering process.

Lack of Accountability
Over the years, few of the many recommendations designed to improve the system and ensure the quality of the nation's statistics have been implemented.⁷The reason is managerial: Responsibility for improving the gathering of statistics is scattered across at least 70 agencies, and no single official or agency can be held accountable for implementing projects (such as improving the measures of the service sector) that would modernize and improve the system across all of these agencies. Moreover, most statistical agencies are regarded as minor appendages of the various Cabinet-level departments, so promoting important statistical issues often is perceived as less important than other high-profile goals and objectives of those departments. The provision of statistics suffers from this lack of leadership. No one individual or agency has been charged with developing an overall statistical research and development agenda; no one is fully answerable to Congress or the American public for the quality of the statistics produced; and no one is acting as a strong advocate for the importance and use of those statistics. Without accountability and distinct responsibility, the quality of federal statistics will continue to deteriorate.

Absence of a National Strategy

Fragmentation and confusion within the current system also mean that resources are expended to collect data of limited public policy interest while other key areas of society are not measured. For example, although users of data from the U.S. Department of Agriculture can access quarterly statistics on the number of goats lost to predators, they cannot find data on the life expectancy of farmers compared to that of other occupational groups. The Environmental Protection Agency, which resists producing comprehensive estimates of the cost of its regulations, collects as many as three overlapping, inefficient sets of data-in different reporting formats-on hazardous wastes and their management. And while federal statistics on vehicle ownership or household plumbing are available by racial origin for every ZIP code in the nation, there are no available statistics on the relative impact of federal taxation on the income levels of different racial and ethnic groups or on the role religious affiliation plays in creating stable and prosperous families.

A Politicized Process

Deterioration in the quality of the census has been accompanied by increased politicization of the Census Bureau's decision-making process. There is much controversy, for example, over whether to use statistical sampling rather than the traditional head-count method in the next decennial census. Critics claim the latter method undercounts low-income and minority populations. The controversy, described by one commentator as "an explosive political event,"⁸pits lawmakers against ethnic and political organizations and regional groups. Intensive lobbying and legislative efforts to influence the methodology used by the Census Bureau have succeeded in politicizing the statistical process, threatening the integrity of the data which the government is constitutionally required to collect.

The Decline of Public Trust
Perhaps the most serious problem facing the federal statistical system is presented by the declining level of public trust in government, specifically in whether or not the government will hold the information it collects in the strictest confidence and refrain from using it in some way against respondents. Protection of the confidentiality of information collected for statistical purposes is fundamental to the gathering and development of high-quality data. The reason: Unless respondents can be certain that the data they give to the government for statistical purposes will not wind up being used against them for regulatory or enforcement purposes, they will refuse to respond to a survey or-if they do choose to respond-will provide inaccurate information.⁹In either case, the statistical series produced will be inaccurate.

Currently, protection of confidentiality in the federal statistical system is neither uniform nor complete because individual agencies within this decentralized structure originated at different times and for different legislative reasons. As a result, the system now operates under a complex set of regulations, executive orders, and laws that differ in their application among the various statistical agencies. At one extreme is the Census Bureau, with protections that prohibit other federal agencies from accessing the microdata even when they paid the Census Bureau to collect it. At the other extreme are agencies like the Energy Information Administration, which has used data collected in enforcement proceedings against respondents. At times, as in the case of the National Center for Educational Statistics, the law spells out clear confidentiality protection, but Congress has stepped in to weaken protection retroactively.¹⁰

These differences in laws and policies among agencies-differences that hamper and, in some cases, prohibit the exchange of microdata-are just as important as confidentiality. Agencies either must forego the use of data already collected or must re-collect the data themselves. The burden on respondents is therefore increased, and many people-believing the government statistical system to be inefficient and incompetent-have chosen not to respond at all to surveys unless required by law to do so. The problem is complicated further by the fact that, under current law, agencies may not undertake comparative microdata research and universe list comparisons and may not work to ensure the uniformity of classifications with data collected by two or more different statistical agencies.

THE BENEFITS OF CONSOLIDATION

According to former Bureau of Labor Statistics Commissioner Janet Norwood, consolidation is "the most effective solution to the problems of the federal statistical system."¹¹The current system has neither the advantages that come from centralization nor the efficiency that comes from strong coordination. Centralization, or consolidation, alone cannot solve all the problems facing the current system, but significant improvement cannot occur without it.

Among the advantages offered by consolidation are:

• Better data at lower cost. Consolidation of federal statistical agencies would reduce the costs of bureaucratic overhead and the production of duplicative, contradictory, and inconsistent data. A consolidated statistical agency could achieve greater economies of scale by reallocating resources to priority areas and encouraging greater mobility and cross-fertilization of knowledge across statistical surveys. Existing surveys could be redesigned to improve cross-tabulation and data matching. This would improve coordination of activities between various agencies as well.

• A coherent national research and development strategy on statistics. A single statistical agency responsible and accountable for adapting the federal statistics system to changes in American society would make a national research and development strategy possible.

• Improved confidentiality and public trust. Consolidation would eliminate the piecemeal tangle of confidentiality protection laws that now exists among statistical agencies. Respondents who supply data to the federal government would have legislative assurance that the information they submit for statistical purposes would not be used against them in enforcement actions. Survey response rates and data reliability would improve; data sharing between the consolidated agencies would improve and reduce public reporting burdens; and the consolidated agencies could ensure uniformity of classifications in the data collected and begin comparative microdata research and universe list comparisons. Strict firewall protections against enforcement agencies, combined with penalties for disclosing or viewing information, would ensure that the creation of a centralized statistical agency is not a threat to the privacy rights and civil liberties of Americans.

• Greater independence. Currently, all statistical agencies are under larger departments or agencies. Even though intentional political bias is not discernible at the major agency level, much of this work is intimately related to the policies and priorities set by the political overseers of these departments. The creation of an independent statistical agency would free these departments and agencies from politicization by either the Administration or Congress and improve the credibility and integrity of the statistics they produce.

HOW TO REFORM THE FEDERAL STATISTICAL SYSTEM

As the 105th Congress begins to discuss how to improve the federal statistical system, it should consider several key principles to ensure that taxpayers and data users receive the greatest benefits from reform. To improve the quality of federal statistics, Congress should consolidate as many statistical agencies as possible under one agency. Guidelines should be structured to improve privacy, ensure confidentiality, and strengthen oversight and coordination. Finally, the new agency should be completely independent so that it can operate objectively. With these principles in mind, Members of Congress should implement the following steps to reform the nation's statistical system:

• Consolidate the four major statistical agencies-the Bureau of Economic Analysis, the Bureau of Labor Statistics, the Bureau of the Census, and the National Center for Health Statistics-into a new Bureau of National Statistics. Although consolidating the four largest statistical agencies into one would eliminate some of the duplication, the largest budget savings and benefits would come from integrating as many of the smaller agencies as possible over a period of seven to 10 years. If the objective-to consolidate the federal statistical system-is clear, a commission could develop the detailed organizational structure for the BNS, specific recommendations for the actual consolidation process, and a schedule to be followed in consolidating the agencies. However, another commission should not be set up to study consolidation of the four largest statistical agencies. Two-thirds of the committees or commissions that have looked into this issue over the past 30 years have recommended some form of consolidation.¹²

The Bureau of National Statistics should have two major components: a public statistics division and a government statistics division. The public statistics division, charged with producing statistics of general public interest, initially should contain four bureaus that incorporate the functions-Census and Demographic, Labor and Prices, National Income, and Health-now carried out by the four major statistical agencies. Additional bureaus could be added as other statistical agencies are combined within the BNS.

The government statistics division could operate in a manner similar to Britain's highly regarded Government Statistical Service.¹³It would include a corps of statisticians who work within the various independent agencies and Cabinet departments to serve individual administrative statistical needs. These statisticians would be under the day-to-day control of managers in those individual agencies but would be employed and promoted by the BNS, and therefore mobile across the federal government.

• Improve privacy and confidentiality by developing uniform privacy protection provisions. The confidentiality protection laws that have evolved piecemeal among the various statistical agencies should be replaced with uniform privacy provisions that permit the exchange of confidential information for statistical purposes only. Such provisions should allow data sharing within the BNS for purely statistical purposes and prevent access by agencies or individuals outside of the BNS. To ensure the honest cooperation of respondents, information provided by the public to the BNS should not be-and must not be-used in any enforcement actions against survey respondents. Strict firewalls should be set up between the federal government's statistical and enforcement agencies.

• Improve management and coordination. Responsibility for the management and coordination of the government's statistical activities clearly should be vested in one agency. This agency, preferably the BNS, should be given the authority and management structure to develop an overall statistical research and development agenda and to implement modernization and improvement projects across all agencies. It should be headed by an independent Chief Statistician of the United States. Members of the board of directors should be appointed by the President, subject to confirmation by the U.S. Senate, and serve seven-year terms. They should not be federal employees. The board's responsibilities should include evaluating the system and recommending to the President and Congress the agency's budget priorities and possible survey improvements, as well as overseeing the quality and objectivity of the data produced.

• Ensure independence and objectivity. The two most important attributes of any statistical agency, from the standpoint of assuring its objectivity, are longevity of leadership and independence from political pressure. Fixed terms of more than four years in office would give agency administrators the ability to resist special-interest demands to change how data are collected and interpreted. The length of time in office would have an important effect on the efficiency of the agency's operations-a factor that is critical to the future integrity of the government's statistics.¹⁴

• Strengthen oversight of the federal statistical system. Although the BNS would be charged with responsibility for overall federal statistical policy, it is extremely important that effective oversight be maintained by Congress and the White House Office of Management and Budget to help ensure the independence and objectivity of any new statistical agency. OMB should comment on classification and quality standards, continue to solicit public comments, and review and approve all federal information collections. Congress also needs to ensure that the goals and objectives of public laws are being met, including efforts to minimize the paperwork burden and to provide more useful information to the public.

These steps offer Congress a solid foundation for the implementation of sensible reforms to consolidate the nation's statistical activities, improve the quality of U.S. statistical data, reduce costs, and alleviate the problems that currently plague the system.

CONCLUSION

Although providing accurate statistics has been a basic function of the federal government since 1787, federal statistics occupy a much greater role in the life of most Americans today. Statistics inform individuals as to the overall health of the nation's economy and set the tone for political debate. They shape thousands of business decisions every day and form the basis for policy determinations at all levels of government. Hundreds of billions of dollars are taxed, spent, and transferred each year based on the data produced by the more than 70 statistical agencies within the federal government.

However, the federal statistical system is in jeopardy. Its problems have been documented by the government's own watchdog groups, and it labors under an organizational structure that prevents it from implementing solutions to these problems. Congress should act immediately to develop a new structure through which the federal statistical system can fulfill all of its constitutional and legislative mandates.

The gains to be realized from improving the efficiency of the system are far greater than mere budgetary savings that come from reducing overhead or combining surveys and data. The development of a coherent national statistical research and development strategy, accountability, an ability to resist politicization, closer coordination of statistical surveys, assurance of confidentiality, and restoration of public trust all are benefits that would improve the quality of federal statistics used by policymakers, community activists, and business leaders alike. Over time, these public and private gains from an improved, efficient, and reliable federal statistical system will prove that Congress's actions were justified.

Endnotes

¹ Office of Management and Budget, Executive Office of the President, Statistical Programs of the United States, issues for Fiscal Years 1980-1994.

² In 1995, the Statistical Policy Branch had a staff of five: a chief statistician and four other professionals.

³ For example, the National Agricultural Statistical Service was created by Congress in 1862; the Bureau of Labor (now the Bureau of Labor Statistics) was created in 1884; and the Energy Information Administration was established in 1977.

⁴ Toward a More Accurate Measure of the Cost of Living, final report to the Senate Finance Committee from the Advisory Commission to Study the Consumer Price Index, December 4, 1996, p. 7.

⁵ U.S. General Accounting Office, Decennial Census: 1990 Results Show Need for Fundamental Reform, GAO/GGD-92-94, June 9, 1992.

⁶ U.S. General Accounting Office, High-Risk Series: Quick Reference Guide, GAO/HR-97-2, February 1, 1997.

⁷ See Joseph W. Duncan and Andrew C. Gross, Statistics for the 21st Century (Chicago, Ill.: Irwin Professional Publishing, 1995); see also U.S. General Accounting Office, Statistical Agencies: Consolidation and Quality Issues, GAO/T-GGD-97-78, April 9, 1997.

⁸ James Glassman, "A Virtual America?" The Washington Post, May 13, 1997, p. A17.

⁹ For example, census mail response rates have been falling since the Census Bureau first implemented its mail-back approach in 1970. If past trends continue, the Bureau believes the mail response rate could decline from the 65 percent rate achieved in the 1990 census to 55 percent in the 2000 census. Besides an increasing distrust of government, the decline of survey response rates has been attributed to, among other things, the increase in dual-income and single-income families (no one is home to respond to the survey) and a substantial increase in "junk" mail (causing the public to be more inclined to throw out mail survey forms immediately). From U.S. General Accounting Office, 2000 Census: Progress Made on Design, But Risks Remain, GAO/GGD-97-142, July 14, 1997.

¹⁰ Janet L. Norwood, Organizing to Count (Washington, D.C.: Urban Institute Press, 1995). Janet Norwood was Commissioner of the Bureau of Labor Statistics from 1979 to 1991.

¹¹ Ibid., p. 71.

¹² Ibid., pp. 7-24.

¹³ Britain's Government Statistical Service (GSS) was established in 1993 as part of a series of reforms. For a brief overview of the British system, see http://www.statsbase.gov.uk/gtos2/dbguide.htm.

¹⁴ Norwood, Organizing to Count, p. 81.