Politics and Federal Statistics^*

^*This lecture, supported by the University of Connecticut and the Pfizer Company, is the fourteenth in the series selected by the American Statistical Association's program for the filming of Distinguished Statisticians. This is a transcript of the speech, which was given on April 24, 2000.

The federal statistical system in the United States regularly compiles data on the issues that concern the American people, data that play an important role in the lives of our citizens. Although a very small proportion of the government's budget is spent on statistical information, these data are an integral part of the functioning of our democratic system of government. Knowledge is necessary for effective policy formulation, and statistical measurement, a part of that knowledge, is essential to our future. That measurement must be carried out in an environment that values the technical competence required to open new frontiers and to undertake seminal research. Nowhere is this more important than in the field of public policy, where the issues to be defined and described have become increasingly complex and the actions based on the knowledge gained from statistical information have increasingly large effects.

As a result, the statistical system in a democracy has a heavy responsibility. The data it produces must represent what has actually happened and not what people believe should have happened. The data must be objective because they will be used to evaluate government effectiveness, and they must be available to all citizens on an equal basis.

This year, as we begin a new Presidential election campaign, we must remember that government statistical information will be used to estimate the number of people living in poverty, evaluate the success of welfare reform and crime reduction, monitor our health and the problems of our environment, and to assess our educational progress. Under these circumstances, we should not be surprised to find that, from time to time, the President, his Cabinet, or the opposition party in the Congress would like to develop ways to control a statistical series perceived “misbehavior.”

Summary

The discussion that follows is based on my own experience over the last 30 years at the Bureau of Labor Statistics and in the federal statistical system. I will start with the political controversy over the release of BLS unemployment estimates in the early 1970's, then move on to the more recent issues of criticisms of the Consumer Price Index, and finish with a discussion of some of the strong political pressure on the 2000 census.

The Politics Surrounding the Release of Statistical Data

For some 20 years before 1971, the professional heads of the Bureau of Labor Statistics’ unemployment and price offices customarily held briefings to explain their data to the press. These press conferences were discontinued early in 1971, in the midst of political controversy over the Bureau's interpretation of an unemployment rate change.

The problem began when conflicting interpretations of the jobless rate for January 1971 were given to the press at the same time – one by Secretary of Labor, James Hodgson, and the other by BLS Assistant Commissioner, Harold Goldstein. The January unemployment rate, at 6.0 percent, was two-tenths below the level for December. The Secretary said that the drop in the jobless rate was “very significant.” At just about the same time, in the BLS press briefing, Goldstein referred to the change as “marginally significant.”¹ Goldstein was cautious because the jobless rate had been rising for some months before January, the change of two tenths of a percentage point was barely statistically significant, and he did not want to declare a new trend based on only a single month of data. Secretary Hodgson, on the other hand, was anxious to project the political value to the Administration of a declining jobless rate. A conflict of views occurred again in March, when Secretary Hodgson spoke of a further drop in the rate as “heartening” whereas BLS pointed to “mixed signals” because both the number of people employed and the hours they worked also declined.

Newspaper reporters focused on the differences in interpretation, and some charged the Administration with attempting to politicize the jobless data. In mid March – the Secretary of Labor announced the termination of the BLS press conferences on the unemployment and consumer price releases. He wanted, he said, to “…avoid the awkwardness of subjecting the professional staff of the Bureau of Labor Statistics to questions with policy implications.”²

The media insisted that the BLS briefings were important, because reporters used them to get an objective view of data developments. Congress and the professional community began separate reviews of the charges. Senator Proxmire, then chairman of the Joint Economic Committee of the Congress, asked the BLS Commissioner to appear before the Committee to explain the data and announced a new regular process of Committee Hearings on the day of the release of unemployment data.

A few months later, in July of 1971, the BLS press release for June reported a significant drop in the jobless rate and warned that the decline might be overstated because of technical problems with seasonal adjustment. The White House was reported to have made its unhappiness clear. The press again charged the Administration with attempts to muzzle the people at the BLS.

Then the Administration took stronger action. BLS Commissioner Geoffrey Moore announced a change in several senior positions at the BLS. Two members of the senior staff–one of them Harold Goldstein–were transferred out of jobs dealing with sensitive data, and a new Deputy Commissioner for Data Analysis was appointed from outside the Bureau.

Although Moore insisted that he merely wanted to improve the BLS, a Washington Post editorial accused the Nixon Administration of filling the Bureau with political appointments.³ The Congress scheduled hearings; a GAO investigation was ordered, and two professional organizations established a joint Special Committee on the Integrity of the Statistical System.⁴ The Social Science Research Council began yet another investigation of the politicization issue.

In the midst of this discussion, Secretary Hodgson began to participate in the BLS meetings on the preparation of the price press releases. Never before had the Secretary been involved in review of these releases before they were issued. I was then the Chief of the Consumer Price Division and participated in those meetings on the CPI in the Secretary's office. The BLS staff held its ground in these meetings, but it is important to note that they took place.

In preparation for this lecture, I wanted to determine whether the White House itself had been involved in these changes at the BLS. I went out to the National Archives to listen to tape recordings of President Nixon's conversations on tapes that had recently been opened to the public. I was surprised–and very much distressed–to find that the President, himself, was not only involved in the discussions about the BLS, but, in fact, had directed his staff to reel the Bureau in.

In referring to the Bureau of Labor Statistics staff and Harold Goldstein's press briefing, the President said–and I want to emphasize that the following excerpts quote his exact language:

Nixon: “I know these little bastards. We have a bunch of people in the bureaucracy to screw us. Here Jim Hodgson goes out and does a very good job, and they make him look like a fool. And we have [Commissioner] Moore, an ass.”⁵

There were a number of conversations about the BLS. In talking more specifically about the unemployment rates, President Nixon said:

“We all know about statistical aberrations. Why is it that when it went to 6.2 and 6.3 they did not say it was a statistical aberration? This is a deliberate job so we don't get any credit when the rate goes down.”⁶

The BLS “problem” was discussed in several White House conversations. In one of these conversations, Mr. Haldeman, said to the President:

“We just have to get rid of him [Moore] and put a political person in… We can reorganize the bureau. In that process, the people we don't want will quit.”⁷

The President's views in these conversations were quite clear:

Nixon: “We are not going to wait… They just say it is statistical aberration; they cut you to pieces. The newsmen are against us. I don't want to jimmy the facts, but they screw us at every release.”⁸

Indeed, the President wanted quick action. He said:

“Question is how to remove Goldstein … a way that we had talked about this morning is to say BLS is in need of reorganization and put all statistics in the same box together with the census bureau…We can rearrange it to assign Goldstein so he is not in charge of the figures…. We have to have control of the releases. We just can't have them go out and cut us up.”⁹

The Bureau's comment about seasonal adjustment problems with the June unemployment decline made the President even more furious. The issue was discussed on the Nixon tapes for July 2, 1971:

Nixon: “I have had it. I don't want any crap this time. They screwed us on this one. He [Goldstein] is to take a polygraph test on it tomorrow. If he did it, he is out of the government…¹⁰ “He was ordered not to do this. He is a Civil Service employee, but we can move him. He is an all out son of a bitch who has been against us for 20 years. We have to be tough on this. We are running the government.”¹¹

It seems clear from the Nixon tapes that the President wanted to control the interpretation of data compiled by BLS. Although he realized that it was probably not possible to change the numbers themselves, the White House tapes show that he was intent on controlling comments about the data that were given out to the press.

Without the benefit of the Nixon tapes, however, the Congressional Subcommittee that investigated the controversy reported that its staff could not find “…any evidence to demonstrate the allegations of politicization of statistics.” They went on to say:

“This was probably due to the high regard held for the Government professionals associated with the gathering, assembling, and reporting the statistics and the integrity of the professionals which has been built up over the years.”¹²

Changes in the Release of Economic Statistics

In an effort to quiet the controversy, the Department of Labor issued new rules that focused on the BLS statistical role, but no real change came until the early days of the Ford Administration, when the Council of Economic Advisors considered how best to restore confidence in the objectivity of the government's release procedures. The driving force for this initiative was Sidney Jones, then a member of the CEA. I participated in those meetings to reestablish trust that Sid chaired.

The Committee's solution was contained in a new OMB Directive for the release of the government's major economic indicators. With only minor additions, those new procedures have remained in place for almost 30 years. Release dates are published in advance. The statistical agency provides the data to the Chairman of the Council of Economic Advisors on the afternoon before the release so that he can inform the President. No one else, not even the Secretary of the Department in which the data were compiled, receives the data from the statistical agency until the next morning when they are provided to the press and to the public. In addition, the new rules provided that no political policy official may comment on the data until they have been in the public domain for at least one hour.¹³

This system for the major economic indicators has worked well, rarely having been broken even during the confusion of changes in administration. I know that it worked at the BLS because, as Commissioner under three Presidents, I had to brief six new Labor Secretaries about the order and explain when they could receive the data we issued and when they could comment on them.

The hearings before the Joint Economic Committee replaced the BLS press conferences. Over the years, the hearings provided an opportunity for the public and the press to hear the BLS objective view of the data as well as the more political interpretations by Committee members from both major political parties. Some 14 years later, at a Joint Economic Committee Hearing, Senator Proxmire, who had initiated the unemployment hearings explained why he had done so. He said:

“I go through this review to remind us that the unemployment data constitute political dynamite. The unemployment rate is the big political enchilada in our country especially to an incumbent President and the members of his party in the Congress. The most obvious practical proof of the wisdom of a President's economic policies is what those policies do to jobs…. When unemployment rises, political discontent grows.”¹⁴

The Reagan Period

The system continued to work well during my tenure as Commissioner of Labor Statistics. But there was one episode that occurred at the beginning of the Reagan Administration that I think is important. I had not met our new Secretary of Labor, Ray Donovan, before I went to his first staff meeting. I listened as he introduced his new Chief Information Officer and said: “From now on, I want all communications with the media to be made through the Information office.”

That would mean that BLS would be cut off from all contact with the press and would not be able to discuss the statistical aspects of data with them. So, I said: “Mr. Secretary, you mean all contact except for the Bureau of Labor Statistics.” But he replied: “No. I mean all of the Department.” I asked if we might talk about it after the meeting. I can tell you that I don't remember anything that happened during the rest of that meeting–I was trying so hard to think of how to handle the issue with the Secretary.

After the meeting, I said to him: “You know, the media is used to talking to the BLS staff.” I knew from press reports that Secretary Donovan had deep admiration for the President, and so, I said: “Mr. Secretary, if we cut off contact, people will blame the President. He will be accused of politicizing the interpretation of the data, and he will be quite innocent. That would be unfair to him. Secretary Donovan thought about that and, after a moment, said: “I guess you are right. But I still don't like it. Let's just say that I trust you. You can have contact with the press, but no one else at BLS should do so.” But I replied: “Mr. Secretary, that just wouldn't work. Why don't we just say that I will take full responsibility for anything that anyone at BLS says to the media.” And he said, “OK.” Of course, I had responsibility for what happened at BLS anyway, and I knew that the staff could be trusted to deal with the media in a professional and objective manner.

I have often thought about that meeting and how important it was to take a stand on the issue. Interaction of professional staff with the media has been important in much of the discussion on statistics that has taken place in this country. Scientific design and estimation issues increasingly have become interwoven with the policy purposes for which data are used. I believe it is important for the media to understand and write about the statistical aspects of the data.

That is what has been happening in the discussion of Consumer Price Index issues and, in many ways, of the 2000 Census as well. Both have been under considerable attack by the Congress. The first, the controversy over the Consumer Price Index involves issues of concept whereas the second, the controversy over the Census involves issues of methodology. But the politics of the numbers is the major issue driving the discussion in both cases.

The Consumer Price Index

For some time, especially in the last few years, the Consumer Price Index, the nation's best-known inflation indicator has been the focus of political controversy, indeed even of attempts at political interference.

The country had for many years run a sizeable deficit, and some politicians saw the CPI as an important engine of the nation's rising debt. The index is used to escalate wage and social security payments and to index income tax brackets and other entitlement programs. As government outlays for many entitlement programs went up, Administration and Congressional officials became more concerned about the cost of indexation.

Given the many uses of the CPI, it is not surprising that political officials of both parties saw changes in the index as the way to solve the country's budget deficit. Politicians–both in the Congress and the Administration–began to realize how the index affected both sides of the government's balance sheet–money spent and money received.

The Recent CPI Controversy

The most recent controversy over the CPI, erupted about five years ago when the CPI became embroiled in the deficit fight. The discussion started when Federal Reserve Chairman Alan Greenspan told the Congress that the CPI overstated the cost-of-living and suggested that the CPI, when used for indexation of federal government entitlement programs, should be reduced by one percentage point. Speaker of the House Newt Gingrich took a different view. He was quoted in the Washington Post¹⁵ as having said:

“We have a handful of bureaucrats who, all professional economists agree, have an error in their calculations. If they can't get it right in the next 30 days or so, we zero them out…” transfer the job to the Federal Reserve or the Treasury, …”and tell them to get it right.”

Several Democratic Senators took issue with Speaker Gingrich against “…pressure exerted by politicians” and in favor of …”careful study.”¹⁶

The issue was not quite so simple as Mr. Gingrich indicated. Although many economists believe that the CPI does overstate the actual cost-of-living, they do not agree at all on the amount of the upward bias in the index. They also do not agree on what a true cost-of-living index is, nor do they agree on exactly how such an index should be calculated.

It is important for this discussion to understand exactly what the CPI is and what it is not. The CPI measures the change in the prices of a market basket of goods and services people buy for consumption. The index has a complex design of interlocking probability samples based on a number of surveys. There is a sample of goods and services—the market basket—and weights for each of the items in the basket to reflect their relative importance. Prices are collected for these items in a sample of stores and other retail outlets in a sample of statistical areas (CPI “cities”). The specific variety of the item is selected through probability sampling in each outlet, and the item is carefully defined–with an individual description or specification–so that the pricing agent can collect the price for the same item from one month to the next.

The recent public discussion of the CPI reflects some confusion over differences between an index based on the price of a fixed market basket in two different time periods and a cost-of-living index that would measure the cost of a market basket in one time period that provides a consumer with the same satisfaction as the market basket of the previous time period. The issues in the debate on the CPI usually focus on five major areas: first, substitution bias (when consumers change purchases in response to the relative change in the prices of items), second, the difficult problems in adjusting the priced items for changes in quality, third, outlet bias, fourth, methods for introducing new products into the index, and fifth, the frequency with which the underlying expenditure weights should be changed. There are also a series of difficult methodological issues such as the use of geometric vs. arithmetic means for combining prices, alternative approaches to sample design, difficult-to-measure products or services, and use of superlative vs. Laspeyres formulae.

The Boskin Commision Report

The Senate Finance Committee held several hearings on allegations of “upward bias” in the CPI. The Committee, intent on considering an adjustment of the CPI for use in indexation, established an Advisory Commission to study the CPI under the chairmanship of Michael Boskin, former Chairman of the Council of Economic Advisors. I testified at one of the Finance Committee Hearings and made clear my view that while the CPI could always be improved, we should not assume that all index problems resulted in upward bias; some, undoubtedly, result in lower bias as well. The real problem, I said:

“is not how to adjust the CPI but whether the country can afford the amount of indexation provided by law. …Congress should not attempt to solve a political problem by legislating the level of a statistical index. I can think of nothing …more damaging to the nation's statistical system.”¹⁷

Unfortunately, some Congressmen and Senators seemed ready to legislate methodology for the index. One proposal went so far as to provide that the Bureau would receive its appropriation only if it lowered the CPI.

On December 4, 1996, the Boskin Report was issued.¹⁸ The Commission estimated that the CPI overstated the cost of living by about 1.1 percentage points and that such an overstatement “would contribute about $148 billion to the deficit in 2006 and $691 billion to the national debt by then.”¹⁹

This is not the place for a detailed discussion of the Advisory Commission's recommendations and the BLS responses to them. Suffice it to say that the Commission made a series of recommendations to the BLS to make changes in the CPI, which, the Commission felt, would make it a better estimate of the cost-of-living.

Reaction to the Boskin report was vocal. Many, especially in the Congress, insisted that the index was erroneous, was causing problems in budget deficit reduction, and that the BLS should quickly change the way in which it calculated the CPI. Others, who felt that the report had oversimplified the difficulties of producing a cost-of-living index (COLI), urged the BLS to continue research and testing on the issues before making large changes. A few expressed concern that the Boskin report focused only on methods for reducing the CPI and paid little attention to changes that might result in index increases. Still others insisted that the discussion was an effort to apply political pressure on the Bureau of Labor Statistics. In response, Senator Moynihan, who had been instrumental in appointing the Boskin Commission, spoke to the Senate about “…policies based on inaccurate statistics” and insisted that criticizing the index did not constitute political interference with the BLS. He said further:

“It is important that our Bureau of Labor Statistics, which is comprised of many superb professionals, even so, be humble enough to recognize that it may not be the repository of all expertise on this subject. There are other views, and they need to be considered carefully by the BLS. Commissioner Abraham would do well to be mindful of this in her second term.”²⁰

The BLS responded carefully to the Boskin Report. They agreed that measurement of the cost of living was a good conceptual goal, but indicated that no one had yet solved many of the practical difficulties involved constructing one. The Bureau pointed to a series of changes that had already been implemented and emphasized that their research was continuing.²¹ The BLS pointed out that, as a government statistical agency, it must use methods that are “…objective, reproducible, and verifiable;” the Bureau could not focus only on judgment but needed hard evidence to ensure the validity of the steps it took to improve the index.²²

It seems clear that, while most economists believe that the CPI is an upper bound on the cost of living, there is no agreement on the exact steps that need to be taken to create a true Cost-of-Living Index. The problem, of course, is that taking steps that would actually improve the CPI–as opposed to simply changing it by some arbitrary estimate–is a complex, expensive, and time-consuming process.

The difference in the recent fight on the CPI and earlier criticisms is that previously the political fire came when the country was experiencing double digit inflation, and the index was considered by many to be too low. Compilation of the index involves many complex conceptual and methodological issues. The problem for the Bureau of Labor Statistics has been that while it is not hard to be in favor of deficit reduction, producing a Cost-of-Living Index can be much harder.

To lighten the discussion of these difficult issues, I would like to quote from a tongue in cheek column written by two Amherst professors entitled, “Don't Let Mere Numbers Stand in the Way of Our Pressing Social Problems.”

They said that the country had “…just won a trillion dollars” from the Boskin report and congratulated the Commission for “…the startlingly simple idea of re-examining constants long held as sacrosanct.” Speaking of the need for flexibility in a world of physical constants, they suggested shortening the number of minutes in an hour by 1.1 percent, thereby saving us an extra 15 minutes a day and reducing the “…inflated value of the gravitational constant … “At the time it was set, men wore breeches and had time to sit around and watch apples fall; … By readjusting the rate at which objects fall by a modest 1.1 percent, the nation could save billions of dollars each year.” Americans, they suggest, “…must finally overcome their vestigial wariness of numbers…We have nothing to lose but our numerical chains.” ²³

It is clear from the case of the CPI–and we shall see as we move on to discuss the decennial census–that the data the government produces has increasingly been enmeshed in the politics of the numbers. People tend to believe the numbers if the numbers support their view of reality, or what they would like reality to be, and to question the accuracy of those numbers which do not support the policies they seek.

The Controversy over the Decennial Census

Even more than the CPI, the decennial census has been embroiled in political controversy. From the very beginning, in 1787, when the decennial census was established to apportion the seats in the House of Representatives, the census was the political compromise for the apportionment of political power among the states. Regular measurement of the population became a clear necessity as changes occurred in the size of the population, its distribution within the country, and new states were admitted into the Union. Shifts in political power among the states caused controversy, but in most cases, the fights among partisan interests involved disagreements on the use of the results of the count.

The system worked reasonably well then, but, more recently, the emphasis has changed. As the plans for the 2000 census were being developed, the politics of the Census shifted from arguments about how to use census counts to challenges of the detailed methodology for collecting the data. The partisan battles have spread to the detailed decisions that had previously been made by the Census Bureau's professional staff. At the same time, the census has become deeply embedded in political disputes among different groups of stakeholders. As the current Director of the Census Bureau has often pointed out, the battles over census methodology have become grounded in the politics of power, the politics of race, and the politics of money.

The population census affects the balance of political power in our country because reapportionment shifts political power among the states, as some states gain, and some states lose seats in the House of Representatives. Because the census provides race and ethnicity data required for implementation of many of our important anti-discrimination laws, it has become embroiled in race and ethnicity political issues. Moreover, use of census data to distribute large amounts of federal money creates partisan divisions among the winners and losers who receive the funds.

It is important to recognize that the census, the largest and most expensive of all our statistical operations, is a massive operational effort. The census must count all residents of the country and place each of them at a particular geographic address. Even with the most vigorous efforts possible, however, some people are erroneously enumerated, and some people are completely missed. The problem is that the undercount is disproportionately distributed, being most prevalent among racial and ethnic minorities and immigrants. And here is where the politics takes over. Each of our political parties has a different concern. Many believe that the Republicans, worried that the undercounted, who tend to be low-income blacks and Hispanics, are more likely to vote Democratic, and tend to oppose adjustment for the undercount. The Democrats, on the other hand, perhaps believing that the uncounted group might, if better represented, vote for them, favor adjustment.

In addition, the increased use of the Census counts brought more political focus to the census. For years, legal battles over the 1980 and 1990 censuses had continued in the courts. Politicians focused on the effects of the numbers on particular programs and legal outcomes, whereas professional statisticians focused their attention on proposed methods for adjusting the census.²⁴

The Census Bureau announced its new plan for 2000. The questionnaire would be more user-friendly; scientific sampling instead of attempts to follow up all those who had not responded by mail would be used. A large, independent post-enumeration survey would be conducted during the final stages of the census process to estimate the number and characteristics of those who did not respond to the census and to provide data required for adjustment of the final census counts. A form of systematic double counting, Dual System Estimation, was to be used as the method for combining information from the survey to measure the undercount and adjust for it. The Bureau announced its plan to publish only a single set of official census counts adjusted for the undercount, hence a one-number census.

Republican Congressmen opposed the plan. Arguing that the new plan would make the census results dependent on staff judgment and that statistical adjustment could not be accurate at the local level, they insisted that the Bureau continue to plan for a traditional census operation. The Census Bureau commissioned three reviews of their research by panels of experts at National Academy of Science. The panel reports, while not approving every part of the Census plan, generally recommended a reengineered Census, sampling for non-response follow-up, an independent post enumeration survey, measurement of the undercount, and statistical estimation of its size.²⁵

The political pressures on the Census Bureau became more direct and more intense. Several members of the Republican majority in the House said that the Congress should refuse to appropriate the funds requested by the Census Bureau for the new plan. The American Statistical Association, concerned about attacks on statistical sampling appointed a “blue ribbon commission” chaired by John Rolph. The commission pointed out that “…The use of sampling … is consistent with sound statistical practices.”²⁶

The politics of the census continued to ferment as the Administration argued for the Census plan, and the Republican majority in the Congress argued against it.²⁷ A letter sent to the Congress by more than a dozen Past ASA Presidents raised “serious objection” to an appropriation provision that sought to prevent the Census Bureau from using “statistical sampling or any other statistical procedure” in the census.²⁸

Finally, the House Speaker and the President compromised on the census appropriation bill. They provided some funds for the Bureau to continue its work and agreed that court review of a proposed legal challenge to the use of sampling in the census should be expedited. In addition, they established an eight member Census Monitoring Board to oversee the work of the Census Bureau.

Speaker Gingrich, along with the Southeastern Legal Foundation filed suit in federal court attacking the legality of the Census Bureau plans for the conduct of the 2000 Census. The Administration filed a motion to dismiss, supported by the House Minority leader and five other House Democrats as well as by the city of Los Angeles, the state of New Mexico, and a number of other municipalities. The Supreme Court heard the case on appeal.²⁹

Meanwhile, as attacks on the Census Bureau's efforts continued to be reported in the press, Senate Governmental Affairs Committee Chairman, Fred Thompson, threatened that Congress would limit funds for the census “…if the accuracy of the results are in question…” Defending the Bureau, the ranking Democrat on the Census Oversight Committee, Carolyn Maloney of New York, blamed the GOP-controlled Congress for having “kept [the Census] dangling on a string.”³⁰

In the midst of all the controversy, on January 15, 1999, the Supreme Court ruled that the legislation establishing the procedures for conducting the census prohibits the use of sampling for the apportionment of House seats.

The Supreme Court ruling, however, did not end the political dispute. The Court ruled only on the portion of the Census Act prohibiting the use of sampling for allocation of House seats among the states. The Court did not rule on the remainder of the language in the same law, which in section 195 provides that the Secretary “shall … authorize the use of” sampling for other purposes “if he considers it feasible.”

As a result, the political debate on the census has moved undercount adjustment from apportionment to redistricting. While recognizing that apportionment of seats must be determined with numbers arrived at without the use of sampling and, therefore, without adjustment for the undercount, the Administration has made clear that it expects that adjusted numbers will be used for redistricting. In fact, the Justice Department notified states that it will use adjusted Census numbers to judge whether redistricting plans hurt minorities, even if the states themselves do not employ those numbers in writing their plans.³¹

The chairman of the Census Oversight Committee, Dan Miller, began another crusade to improve the census–this time by proposing a series of bills to legislate census operations. Arguing that the Census Bureau had spent too much time on plans for sampling and adjustment and insufficient time on ways to improve the census count, he and others in his party introduced bills to require that the Census questionnaire be printed in 33 languages rather than the six planned by the Bureau, that the Bureau spend more funds on advertising, that the census include a count of Americans living abroad as private citizens, that local areas be given an opportunity to review the counts after the collection as well as before, and that the Census Bureau be required to send each non-responding household a second form. The Administration opposed each of these bills, and none has thus far been passed into law.

As though this host of political and legal problems had not caused sufficient problems, members of the Congress began publicizing complaints from their constituents’ about what they regarded as the intrusiveness of the census long form. The long form, sent to a sample of households, includes questions on such matters as housing, work, disability, and income. The long form data is required for implementation of several government programs and for many important statistical indicators.

Majority Leader Trent Lott was reported to have urged people to fill out the basic information, but “if people feel their privacy is being invaded by those [additional] questions, they can choose not to answer those questions.”³² Others in the Congress said they would fill out only the questions “that …make sense.”³³ And, Senator Chuck Hagel of Nebraska gave advice to people to “…fill out what you need to fill out, and don't fill out anything you don't feel comfortable with.”³⁴

Defenders of the form emphasized the importance of the data for government purposes and pointed out that the Congress had had the opportunity to comment on and remove questions two years before, when the Census Bureau sent members the list of questions for review. “If they had a problem with it, they could have raised it then … Don't they read their mail?” asked Representative Carolyn Maloney.³⁵

Response to the long form became even more a political issue when the GOP candidate for President, George W. Bush, said: “I can understand why people don't want to give over that information to the government… I'm not so sure I would do it either.”³⁶ A bill was even introduced to rescind the law permitting the Census Bureau to fine those who did not fill out the long form, but the bill did not pass the Congress.

In response, House Minority Leader Richard Gephardt, said he thought the GOP were “worried that too many people are filling out the forms…They don't want more people counted because they think it serves their political purpose.” Census Director Ken Prewitt expressed concern over a backlash effect on the census response rates. “To tell them not to complete the form is to tell them to break the law,” he said. And in the week or so after the criticism arose in the House, the number of long forms returned was considerably lower than expected.³⁷

The Bureau's latest figures show {as of about April 20) that overall, nearly 65 percent of all households had returned forms to the Census Bureau. The highest counts were in the upper Midwest, the lowest in the South, and cities rank lower than suburbs. The concern, however, is that the gap in response between the short form and the long form was nearly twice that of 1990. Only 54.1 percent of the long forms but 66.6 percent of the short ones had thus far been returned.

And so, the saga of the politics of the census continues. Unfortunately, we can now see quite clearly how the politics of the census has moved into new areas. It used to be that in the early days of the Republic, the census numbers were used in a more or less automatic manner in formulas to allocate House seats. Now, however, the census numbers are used for such things as defining district boundaries, determining the equity of social policies, helping to enforce the Voting Rights Act, and to allocate billions of dollars. So, the political debates have shifted from how to use the counts to the methods used in the compilation of the census itself. This is a serious change, one that threatens the very scientific approach to measurement.

Concluding Thoughts

Our country is a country of change–in the demographic makeup of our people, in our opinions, and values, and in our attitudes toward our civic responsibilities. We are also a country that takes great pride in our free, democratic system of government, a system that can only function well when its citizens have access to data that are accurate, relevant, and objective, for judging the performance of our public officials. We should be concerned but not surprised to find that our elected political officials place a high degree of importance on policy-relevant data and, even that, at times, efforts are made to control the numbers.

It seems clear that the release of statistical information in unbiased and objective form is an essential element of a democratic society. We have seen how efforts were made to control the release of information, how the President himself attempted to influence data interpretation and to use a political litmus test for the personnel of a major statistical agency. The important point to remember, however, is that it did not work. President Nixon tried to control the professional staff at the Bureau of Labor Statistics, but they withstood his demands; they continued to carry out their professional tasks with objectivity. Moreover, the Congress, in the form of the Joint Economic Committee, stepped in to ensure the stability and independence of the system. The Nixon era problems resulted, in fact, in a strengthening of the system and in the establishment of measures for the protection of its independence.

The politics of the CPI has been a very different issue. It involved legitimate disagreements over difficult issues of concept and their measurement, and the importance attached to any methodological change in such an important indicator. The issues were very much affected by the approach to changes in indexation that differed over the years depending on the political clout of the particular stakeholder groups involved.

The controversy, however, also demonstrated the dangers we face as our statistical indicators become more and more embedded in public policy. At least for some political officials, the dollars to be saved by the federal government if the CPI were to rise more slowly, went a long way toward shaping their positions on the issue. For some–although by no means all–of the policy officials involved, the controversy provided a convenient format for discussion of the CPI's conceptual design as a way to produce a lower index, thereby reducing pressure on the government's deficit by spending less for indexation of social security and other payments.

The controversy over the decennial census is, in many ways, similar to that on the CPI because it, too, is based on perceptions of the political effects of the population count. It is different, however, and far more serious, in that the discussion has focused more on survey methods, data imperfections, and statistical approaches to undercount adjustment than on conceptual purity. Congressional oversight over the Census has traditionally been very strong, and proponents of different views have taken their cases to the federal courts. In addition to all of its other uses, the Census data affect the make-up of the Congress itself. As a result, the decennial Census is inevitably thrust into the political spotlight. The danger we face is that political officials may make methodological decisions that are best left to the trained professionals.

In each of these cases, in different ways, the statistical policy-making role of the Congress has developed and changed. The pendulum has swung away from the scientists and toward the politicians. The fact that statistical series have become so important to the development and implementation of the social and economic policies of this country inevitably creates a temptation to manipulate the numbers for political gain. The best defense against these trends has to be the professional quality and the integrity of those who compile our statistical information. We must keep our data free from political influence. The country needs a strong cadre of professionally trained people with the integrity to stand up for that which they believe.

But one point that I hope this lecture has made clear is that the professionals who compile the nation's statistics must be courageous enough to insist that their own work remains free of political influence, they must refrain from participating in political discussion, and they must insist on following careful, objective procedures for the manner in which their work in compiling the data is carried out.

Notes

¹ Investigation of Possible Politicization of Federal Statistical Programs, Subcommittee on Census and Statistics of the Committee on Post Office and Civil Service, House of Representatives, 92^nd Congress, 2^nd Session, October 5, 1972, p.7.

2   Department of Labor Press Release, March 19, 1971.

3 Washington Post, September 29, 1971.

4 See The American Statistician, April 1973, pp. 1-12 for the report of the Joint Committee.

5 Nixon tapes, Conversation No. 537-2 rev., 2/99, July 2-5, 1971.

6 Nixon tapes, Conversation No. 534-11, Rev. 3/98, July 1, 1971.

7 Nixon tapes, Conversation No. 534-11, Rev. 3/98, July 1, 1971.

8 Nixon tapes, Conversation No. 537-2, rev. 2/99, July 2-5, 1971.

9 Nixon tapes, Conversation No. 535-4, July 2, 1971.

10 Nixon tapes, Conversation No. 6-112, July 2, 1971.

11 Nixon tapes, Conversation No. 6-113, July 2, 1971.

12 Investigation of Possible Politicization of Federal Statistical Programs, Subcommittee on Census and Statistics of the Committee on Post Office and Civil Service, House of Representatives, 92^nd Congress, 2^nd Session, 1972, p. 9.

13 Statistical Policy Directive A-91, as amended, is know named Statistical Policy Directive No. 3 in Federal Register, vol. 80, no. 186, September 28, 1985.

14 Hearing before the Joint Economic Committee, December 7, 1984.

15 Washington Post, January 16, 1995.

16 Associated Press quotation, January 19, 1995.

17 Testimony before the Committee on Finance, United States Senate, June 6, 1995.

18 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

19 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.1.

20 Congressional Record, October 9, 1997, S10826-7.

21 BLS, Consumer price Index: Update of Boskin Commission's Estimate of Bias, February 2000

22 From an earlier BLS document, Measurement Issues in the Consumer Price Index, June 1997.

23 Douglas, Lawrence and Alexander George, “Don't Let Mere Numbers Stand in the Way of Our Pressing Social Problems, Chronicle of Higher Education, vol. XLIII, no. 28, March 21, 1997.

24 See Anderson, Margo J. and Stephen E. Fienberg, Who Counts? The Politics of Census-Taking in Contemporary America, Russell Sage Foundation, New York, 1999.

25 Steffey, Duane L. and Norman M. Bradburn, eds. Counting People in the Information Age, National Academy Press, 1994; Edmonston, Barry and Charles Shultze, eds. Modernizing the U.S. Census, National Academy Press, 1995, and Cohen, Michael L., Andrew A. White, and Keith F. Rust, eds. Measuring a Changing Nation: Modern Methods for the 2000 Census, National Academy Press, 1999.

26 American Statistical Association, Report of the Census Blue Ribbon Panel, September 1996.

27 Sampling and Statistical Adjustment in the Decennial Census: Fundamental Flaws, 14^th Report by the Committee on Government Reform and Oversight together with Additional and Dissenting Views, 104^th Congress, 2^nd Session, House Report 104-821, September 1996.

28 Letter to Members of the Congress, June 11, 1997.

29 U.S. House et al. v. Department of Commerce et al. 1998. The Supreme Court actually heard two cases combined into one: Department of Commerce vs. U.S. House of Representatives and Clinton vs. Glavin. The Court only ruled on Clinton vs. Glavin.

30 From article in the morning Roll Call Online, March 26, 1998.

31 Washington Post, March 27, 2000.

32 Washington Post, March 30, 2000.

33 Representative Robert W. Schaffer, Republican of Colorado, quoted in Washington Post, March 30, 2000.

34 Washington Post, March 30, 2000.

35 Washington Post, March 30, 2000.

36 Washington Post, March 31, 2000.

37 Washington Post, March 31, 2000.