FMC Home Link PBS Program LinkFMC Book LinkViewer's Voices LinkInteractivity LinkTeacher's Guide
  Program Segments LinkInterviews LinkMeet the Host LinkCredits Link

FMC Logo 1
(abbreviated titles)

  Closing of the Frontier
  Scientific Racism
  The Children's Bureau
  Recent Social Trends

  The Great Depression
  The Gallup Poll
  World War II
  Suburban Nation
  Sexual Behavior

  The Feminine Mystique
  The Moynihan Report
  Broken Windows
  Middletown IV
  Census 2000


FMC Logo 2  

Ken Prewitt Interview


Ken Prewitt is Director of the U.S. Census Bureau. He is a former President of the Social Science Research Council, former Director of the National Opinion Research Center, former Senior Vice President, The Rockefeller Foundation, and former Professor in the Department of Political Science at the University of Chicago. 

He is the author of Recruitment of Political Leaders: A Study of Citizen-Politicians. He is a co-author of Social Structure and Political Participation: Development Relationships and Labyrinths of Democracy: Adaptations, Linkages, Representation, and Policies in Urban Politics.

Ken Prewitt

New River Media Interview with: Kenneth Prewitt
Director, U.S. Census Bureau

QUESTION: Talk about the scope of the 2000 Census. 

KENNETH PREWITT: Census 2000 obviously is exciting, since it is the first Census of the new millennium, or the last Census of this century. Census 2000 has been larger than any previous effort. It's larger because we actually tried to design a Census which was owned by the American people, not just by some statistical agency in Suitland, Maryland, but which the entire American public can sort of feel an ownership of, and commitment to, and engagement with. And what really made it exciting to watch was the massive outpouring of civic engagement and civic commitment across towns, cities, villages, across the entire country.

QUESTION: Was this in part because people felt that there were so many linkages to federal programs that they were really getting something out of it?

KENNETH PREWITT: I think one of the issues that we did stress in Census 2000 was - a quip I sometimes used is - "the IRS form in April taketh away; and the Census form in April bringeth back." And it is true that over the last 30 or 40 years we put in place a huge amount of federal funding, and a condition that it is dependent upon a good Census, approaching $200 billion a year now. Since Census data have to last for 10 years, that's $2 trillion. So $2 trillion of taxpayer money is going to come back to communities and population groups, dependent upon the quality of Census data.

But there is a larger story. I think in Census 2000 than just the federal funding. It was also the story of trying to reverse what's really been a three-decade-long decline in civic engagement in the Census. We have watched the mail-back, as we say the mail-back response rate - how many people return their form in the mail - drop every decade by 10 percent just between 1980 and 1990. So we have tried to make Census 2000 a test of the proposition that you could actually reverse civic disengagement. And we are extremely pleased that the final return rate in Census 2000 is better than 1990. We didn't only stop that decline; we slightly reversed it.

QUESTION: Describe the relationship between the mail-back rate and the number you actually count.

KENNETH PREWITT: The mail-back response rate is, from the number of households we sent a form to, how many mailed it back in. And in 1990 that number was about 65 percent; but in 2000 it was about 67 percent. Actually that's an odd number because it doesn't yet take out all of the housing units which are vacant. When you take all those housing units out, which we do later on in a Census cycle, then of course that number increases - it increases in 1990 and it increases again in 2000. We estimate that the 2000 actual participation rate, in terms of mailing in the form, is going to approach 80 percent. That's an extraordinary accomplishment for this society. That's way higher than most mail-out survey responses of course.

QUESTION: Counting the interviewers who go out to get the other 20 percent, what do you estimate the undercount will be, and how does it relate to prior years?

KENNETH PREWITT: The undercount of course, the number of people we simply don't include in the count - it's actually a net undercount, because in a count there is also an overcount. Indeed in 1990 we estimated about 4 million overcounted, that is double-counted people, and about 8 million not counted or undercounted. So the net was about 4 and a half million undercounted. That percentage in 1990 was slightly higher than we had seen in 1980, which had been a reduction since 1970. We have been measuring this systematically really for 40 years now.

The expectation of 2000 is that we have had a good Census in 2000 - operationally in terms of public cooperation rates and so forth. That does not necessarily mean that you reduce that undercount. The undercount is really a very difficult phenomenon. A large part of it actually comes from households from which we get a form, but they don't put everyone in that household down on the form. About half the undercount are kids - kids who are just left off the count form. So that isn't that our operations didn't work - we got to the household, we got a form back. But for whatever reason they didn't put all the kids down. So that's part of what contributes to that undercount.

Sitting here today, we don't yet know whether the 2000 undercount will be less than it was in 1990. We do not know that until after we do a major piece of evaluation work. In 1990 the measured undercount was about 1.6 percent of the population. In 1980, I think 1.4 is what I want to say. But we might have to redo it. The measured undercount in 1990, net undercount, was about 1.6 percent, which was actually a worse undercount than 1980, which was measured about 1.4 percent.

QUESTION: So this whole argument we had in the public discourse - and it was a huge argument - was about whether we are going to count 98 percent of the people or 99.

KENNETH PREWITT: It is true that there is a very large argument going on in this country, a political argument about the undercount, and what to do about it. And though the absolute number seems fairly small - one and a half percent, say - the fact is that one and a half percent is not even across all population groups, all demographic groups. In 1990 it was over 10 percent for Native American Indians living on reservations. For African Americans in central cities, it was 6.7 percent. Different population groups, different rates. So if the undercount were evenly spread across every population group in every geographic area, then you wouldn't have to concern yourself with it. But because it is unevenly distributed across different groups, then it creates obviously a lot of interest and concern, because a group which is undercounted with respect to the representational process of federal funding is going to get less than its fair share of that resource.

QUESTION: Please explain the system for selecting race and ethnicity on the forms.

KENNETH PREWITT: In the preparation for Census 2000, the Office of Management and Budget conducted a quite extensive effort across the entire federal statistical agencies about the issue of how do you measure race in this country. Nothing is more difficult. Nothing is more difficult than measuring race. We don't know what race is. Is it biological? Is it sociological? Is it characterological? Is it linguistic? Is it a skin color? There's as many different anthropological theories and biological theories about what constitutes race as there are people working on this topic. It's an enormously complex and elusive phenomenon. We have been trying to measure it one way or the other since 1790, but as far as I am concerned, never really successfully.

So going into the Census 2000, because of the growing rates of intermarriage in the population across different racial and ethnic lines, it was thought by the Office of Management and Budget and the statistical agencies that we should make allowance for people who are multi-racial.

Now, how to do that? The way we did it in Census 2000 was to have five major race categories: white, black, Asian, Native Indian, American Indian; plus we had Pacific Islander/Native Hawaiian, and then another line that said "other." So even though with in some of those categories you had subcategories, like in the Asian category you could put down whether you were Chinese, Vietnamese, Japanese or whatever, all of those got coded back to Asian. So we are only talking about five major racial groups. Hispanic is not a race, Hispanic is an ethnicity, because you can be Hispanic black, Hispanic white, Hispanic Indian and so forth. So five racial categories.

However, in Census 2000 the decision was to allow people to tick more than one box. That is, if you believe yourself to be part white and part black, you tick both of those boxes. You don't give us the proportion of yourself that you want to put in those boxes; you simply tick both boxes. Or you can tick three boxes. It's called sometimes the Tiger Woods phenomenon. Tiger Woods after all would presumably tick four boxes: Asian, Native American Indian, white, and black, because he sees himself coming from all four of those racial traditions, and he said so. So he would be four categories.

If you take those five basic categories and create all the combinations, including as I say one additional one, which is "other," you come up with 63 different combinations. Now, we don't think the American people will be using all of those 63 in any great numbers. We expect the overall use of multiracial opportunity to be somewhere in the 2 to 3 percent in 2000. We won't know until we report the data of course shortly.

QUESTION: But to do that do you have to add in the Hispanic and all the ethnicities, and those are other combinations?

KENNETH PREWITT: No. What we do - you now have the Hispanic population, the non-Hispanic population. With the 63 categories apiece. So you would have 126 categories if you divided into Hispanic/non-Hispanic. But of course you don't have to do that. You can divide it any way you want to. You can divide it by age. You can say, How many people between 0 and 12 and 13 and 24 and so forth and so on are different kinds of racial characteristics? But the actual racial characteristic measure itself allows for five discrete categories and any combinations of those five categories, which is what produces the 63 separate categories. You get to 126 only by saying you divide it between Hispanic and non-Hispanic. And that's what produces 126. But you don't have to do it that way. That's a choice that the analyst or the commentator could make. As the Census Bureau itself, we're only producing 63 discrete racial categories.

QUESTION: I gather a number of minority groups were not happy with this idea of having a multiracial box, because it would, they thought dilute their influence.

KENNETH PREWITT: In Census 2000, when the data were being prepared for, the multiracial question was being analyzed and thought about, it obviously came to the attention of the agencies of the federal government who have enforcement responsibilities for Voting Rights Act, civil rights laws, nondiscriminatory policies in health, education, voting, housing, labor practices, that how were they going to go about measuring whether discrimination did or did not occur. And they made the decision that across all the agencies that have enforcement powers, they made the decision to combine a lot of the multiracial groupings into a smaller number of groupings, and they did that by simply deciding that there would be room for people who were white, black, white-Asian, white-Indian and so forth, and then a residual kind of multi-racial grouping with everybody else. That was the initial decision.

Then the question is when you are actually trying to administer non-discriminatory social policies, what do you do with someone who says, "I am white-black," or "I am white-Indian," or "white-Asian"? And there what the decision was to assign that person to the category which is discriminated against. That is, we have no discriminatory laws in this society - not because we have discrimination against whites, but because we have discrimination against blacks or against Indians or against Asians, and that's what the laws are designed to do. So the decision of the enforcement agencies is to say, "If you are both of those things, if you declare yourself to be multi-racial; that is, white and black - for purposes of enforcing non-discriminatory policies you will be counted as if you were black." It has nothing to do with proportion of you. We only know if you are white-black by your self-identification.

QUESTION: Suppose a person checks black and Asian - how is he counted?

KENNETH PREWITT: As I understand it - again, this is something the enforcement agencies are doing, not the Census Bureau - but as I understand it , it depends upon the nature of the law that is being litigated in that case. If the law that is being litigated is about a discrimination against Asians, then that person would actually be counted in the Asian category.

[But as far as the Census determination of the racial makeup of the country], the person who checks black and Asian in the Census results will always be black and Asian in the Census results - never anything else. The Census Bureau itself - in none of its tabulations does it change the way any individual so describes himself. The person who checks black and Asian in Census 2000 in Census Bureau tabulations will not be in the black tabulation, will not be in the Asian tabulation. He or she will always be in a separate tabulation, which is black and Asian - always - in the Census Bureau tabulations. In Census 2000 there will be no tabulation which just has the five racial categories alone. There may be a tabulation summarizing a lot of data which has the five major Census tabulations, plus another tabulation that says "all people that said multiracial." But there is no person who declares himself to be two or more races who will ever be assigned to only one of those races in the Census tabulations - under no circumstances.

QUESTION: What do you think of the idea of categorizing 63 varieties of Americans?

KENNETH PREWITT: I think in 2000 - in Census 2000, and not just in the Census but other statistical programs that now have launched, this idea of allowing people to declare themselves to be whatever combination of racial characteristics they wish to be, I think this country turned a major corner, a serious major corner in terms of how we are going to think about race in the future. My personal judgment, not necessarily as the Bureau, as the Census director, but my personal judgment as a citizen, as an observer of American society, I would say that it is going to be extremely difficult to continue to do what we call race-based social policy the way we have been doing it, where you have a discrete number of groups. The more proliferation there is in the number of kind of racial groups, how hard it is to administer those social policies.

Now, this country still has racial discrimination, and we are going to have to work hard on how we can enforce laws against racial discrimination without that kind of nice denominator which we have always had - there are so many blacks or so many whites or so many Hispanics or so many Asians and so forth. But in Census 2000 and other statistical series that will come out of 2000, it is going to be extremely difficult to sustain that particular way of administering our nondiscriminatory policies.

We do not believe that coming out of 2000 that the proportion of the people who will declare themselves to be multiracial will be particularly high, 2 to 3 percent. However, in some parts of the country it will be a lot larger than that. Southern California could be 10, 12 percent. Hawaii is likely to be 25 to 30 percent. These are predictions based upon our best sense, but I can't - I don't know for certain those will be the answers.

QUESTION: Many people have criticized this new way of counting as divisive and separatist. You are saying it is the reverse.

KENNETH PREWITT: I guess that the ideal solution to this would be not to have any racial discrimination. If we had no discrimination in this society against different racial groups, then we wouldn't have to go out and measure race. We partly measure race in order to enforce our nondiscriminatory. On the one hand we do have discrimination in this society, and we have to have some sort of social policies to try to correct for it. And therefore we require some kind of measurement of the racial characteristics.

But at the same time we are moving into a much, much more complicated society with respect to interracial marriages, with respect to cross-ethnic group marriages. And at the same time we have got new groups coming into our country who are wishing to declare themselves an independent race. Arab-Americans for example would very much like to have their own line on the racial category. After all, before 2000 we did not have Pacific Islanders and Native Hawaiian as a separate racial category. And that got put into the Census 2000 because of very strong pressure from the Pacific Islander population group and Native Hawaiians.

One of the most confusing thing about the multiracial item or any measure of race in this society is that once you are in the environment we are currently in, which is a mail-out questionnaire in the Census, people are what they tell us they are. And in the previous censuses, going back before 1950, the enumerator came to the door, took a look at you, and said, "I've decided what you are, and marked you down." But once we went into the mail-out questionnaire starting in 1960, then we can only record - we can only measure race in terms of what people declared themselves to be.

Some may object to the racial distinctions as all [many decline to answer the question]. Yet, we are also of course in a bind, because a lot of scholars, a lot of social commentators, journalists, want to look at the dynamic nature of race in society across history, and how can we fulfill this kind of obligation really to social intelligence if we don't make some kind of measures of race. Just imagine there had been no measure of race at all in Census 2000! There would have been a public uproar: "Why in the world are you not measuring race? You have been measuring it since 1790. We want to know how we are changing as a society, how many people are coming from what parts of the world, what their racial characteristics are."

The measurement of race in the society, any society, but certainly this society, is one of the least successful things that we know how to do. And the problem is two different things are going on. One of which it is about identity. It's about what you want to be or what I want to be, what do we want to call ourselves. But it is also about politics. It's about the fact that a lot of resources, representational resources and financial resources are allocated depending on racial characteristics of this society. So we have got two things going on simultaneously: one, what do people think they are and want to be - what is their identity? And, two, what do they want to be with respect to the politics of race in American society? And those two goals are not easily maximized simultaneously.

QUESTION: What was the original purpose of the Census?

KENNETH PREWITT: I think it's very important for the American people to know the Census was put into the U.S. Constitution to perform a political function, therefore it is necessarily political. And that political function was of course to allocate seats in the new House of Representatives. You have got to go back to 1787 and remember that the big battle in designing the initial Constitution was between the small states and the large states. The small states thought that every state should be represented equally in the new Congress as they had been in the Continental Congress, the Articles of Confederation. And the large states, particularly New York, Massachusetts, Pennsylvania, thought, no, seats should be allocated proportionate to population size. The great compromise of course, the great Connecticut compromise, was a bicameral legislation. In the Senate every state is equally represented, and in the House states are represented proportionate to their population size. Once you made that decision you had to go out and count. And so the first Census in 1790, under the direction of Thomas Jefferson, was to allocate the seats in this new House of Representatives.

The other thing, however, why it is so deeply political as an instrumentality of building a democracy, was the country was faced with a question of, "What are we going to do about all this new territory across the Alleghenies? What are the conditions under which the new territories are going to be included in the union?"

Well, the major decision was they shall come in as free and equal states - not as colonies. We could have colonized timber, mineral rights, huge river system for transportation - a very, very rich part of the country laid just across the Alleghenies. But the Founders in their wisdom said no, as we continue to inhabit and spread across this continent, new territory should come in as new states, equal states - Kentucky, Tennessee, Ohio Valley, so forth. Once that decision was made, you had to go out and count every 10 years to figure out when a new territory had sufficient population size to claim the right of statehood. And during that period, by the way, from 1800 to 1850, there was a very close attention to balancing the pace at which free states and slave states were allowed into the Union. So the decenial Census was quite critical to performing these basic, basic nation-building functions.

QUESTION: Can you have a functioning democracy without good statistics?

KENNETH PREWITT: I would say that what I would call the nation's number system, good statistics, is critical to democracy in two ways. First, simply and most recognized, is effective governance: How in the world can you govern a complicated industrial, advanced industrial economy or now our new knowledge economy without decent information by those people who have to write the laws and administer the laws and the programs? But I would say there is something much deeper about the nation's number system and democracy, and that really has to do with how the people hold their governors accountable. How do they decide if they are being well-governed? Well, they partly hold their governors accountable by looking after all at statistical trends. Every time the American public asks a question about health, education, housing conditions, state of the economy - they are really asking a question that is basically going to be answered with a statistic times series or a social indicator. So I think the rudiments of democratic accountability depend upon a healthy, vibrant, high quality national statistical system.

QUESTION: When did the passion for numeration really come into full flower?

KENNETH PREWITT: There's a scattering of course of statistics going all the way back to the colonial period. After all, the British Crown was trying to run its empire by having some good economic statistics - how well were its colonies functioning and producing new products and goods? And so they were actually doing little censuses all across the colonies, going way, way back.

We put the Census into our Constitution in 1790 for basic nation-building purposes. But it's interesting, even in the very first Census James Madison said we ought to do more than just sort of count the people; we ought to do a measurement of, let us say, the industrial strength or the economic strength of the country. He was not successful in 1790 in getting those kinds of questions put onto the Census. But within about 30 years - 20 years really - 1810, and again more in 1820, we begin to expand the collection of data, manufacturing statistics and so forth, for example in the 1820 statistics. But it's on and off depending on the mood of the country and the mood of the leadership how much data we ought to be collecting in the decennial census.

After the Civil War, of course, with the real growth in the industrial economy, then you begin to get a much, much more professional statistical operation. You don't just sort of do this as a part-time job, by the Secretary of State, but you begin to kind of bring in professionals every 10 years, like Francis Walker, of course, to run the Census. And you get a much, much more extended attempt to collect data. The Census Bureau itself is established in 1902, which sort of signals the beginning of what [I call] "the measured century." That's when you need a professional permanent operation to collect not just the Census data every 10 years, but a much more extensive array of data.

And of course to run the history fast forward, you really begin to have major new scientific breakthroughs in the 1930s with sampling theory. And sampling theory is absolutely critical to the notion of a measured century or a measured society, because what sampling theory does, quite simply, is allows you to talk to a reasonably small number of people and estimate or extrapolate from that characteristics of the entire population. And that brings into play then the opportunity to measure things like unemployment, the Consumer Price Index, housing construction, population movement, language patterns - whatever you want - because you can now do it efficiently and effectively. And it was the Census Bureau of course that pioneered sampling theory in the 1930s.

What you want to remember is in the latter part of the 19th century you have the convergence really of social reform and social welfare efforts with the gradual growth of statistical capacity. So you have things like the measurement of poverty, the measurement of children's employment, indeed all kinds of disastrous employment conditions and health conditions. You have the measurement of settlement houses and the way people lived. So you begin to get the measurement of social phenomena as a way almost to shame the leadership into doing something about them. Indeed, if you go back to the social reformers, both in Britain and in the United States in the Victorian period, you will find that they are some of the pioneers in using statistical information, what we would now call social indicators, to describe the condition of the society or the population that needed attention by the political leadership. This convergence of the social reform movement and the growth of statistics is one of the things that kind of plants the idea that you can actually use statistical time series or social indicators to hold the government accountable for its performance.

QUESTION: Doesn't that also sort of automatically engender a lot of people using data for their own political purposes?

KENNETH PREWITT: When numbers are as important as they are to governance, both from the point of view of people trying to push their agendas to try to bring notice to their concerns, and indeed used by the government to try to govern the people, you do create a pressure if you would to kind of create numbers that are most satisfactory for this, that or the other political outcome.

I think one of the things that certainly happened in American society is the increased use of statistics to argue political points. And therefore what you find is you find people on both sides of a partisan argument trying to use data or statistics to say, "Look, I can prove that with my data that welfare reform is working or not working, that deterrence of crime is working or not working." And both sides will use statistical information.

I think what's really healthy about that is the data least puts boundaries to that debate. You now are arguing about the quality of the information available to you and the inferences you can draw from it. And that's much, much better, I think, than having partisan arguments just about anecdotes, which is the alternative. So at least you would hope that political debate gets a little more reasonable, because you have now put some boundaries around what can be claimed or not claimed. And I think that is a development forward for our democracy. What you have to guard against is the deliberate misuse of the data. It's been said, of course - John Adams said it - that facts are stubborn things. I think you can get a picture of a society in only two ways really, the stories the society tells about itself - its fiction, it poetry, its narrative if you will. And the other way you learn about a society is by its statistical measures of it. And I don't mean to discredit the stories, but the stories very often just become excuses for making political points. But at least the statistics have to be taken into account.

QUESTION: The nativists or scientific racists were among the first to use data in social science measurement to prove their point.

KENNETH PREWITT: Let me talk for a moment about the early 1920s. What some people do not realize is that the Census did its task every year from 1790 to 1920. And after the 1920 Census we did not reapportion for the first time in history. Now, why did we not reapportion on the basis of the 1920 Census? It's because there had been a strong population movement from the rural South to the Northern industrial cities during the First World War period. And when the rural conservative Congress in 1920 saw those data, they did not want to reapportion. Now, one of the reasons they did not want to reapportion is because the cities, the Northern industrial cities, were politically radical. The new immigrant groups from Europe who had come into those cities were after all voting left, or bringing in alien socialist ideas. This was the period of the Palmer raids, of an anti-immigration movement in this country. That converged with the 1920 Census data that said not only that - there's a lot of internal mobility in this country - is moving workers from the rural South up to the Northern industrial cities. And the Congress simply said we are not going to reapportion. We don't want to see power move from our areas to those radical areas of the country. And so they refused to reapportion, and we did not reapportion until after the 1930 Census.

In 1920, or following the 1920 Census, the U.S. Congress actually set aside its constitutional obligation to reapportion. It argued about the data, it argued about the formula. And not until after the 1930 Census did it reapportion.

One of the things that happened, by the way, in this period - up until 1910 every time the population had grown we had expanded the size of the House of Representatives. So it wasn't allocating just a fixed pie, it was expanding the pie. After the 1910 Census, really 1912 with the introduction of I think Arizona and New Mexico were the two new states brought into the Union at that time - it was decided to fix the size of the House of Representatives - do not allow it to grow beyond 435. So 1920 was the first decennial where the Congress was faced with the problem of reapportioning the House of Representatives and not being able to expand the size of the House. And that contributed to their reluctance to see powers shift from the rural South to the Northern industrial cities.

QUESTION: Talk about the rise of the opinion poll.

KENNETH PREWITT: One of the really important developments in the 1930s was the gradual emergence of scientific sampling theory; that is, the idea that you could talk to or get information from a small segment of a population and reliably estimate the characteristics of the entire population. This was a new notion really in population sampling. Of course sampling had occurred before for other purposes, but in terms of population sampling it really begins to emerge in the 1930s, largely at the Census Bureau. Some of the leading statisticians of the time. But of course it was immediately picked up by some of the private sector organizations, like the Gallup organization, which wanted to use this new technology to really begin to do public opinion surveys.

The Census Bureau of course itself was not interested in public opinion surveys; that's not what its job was. It first used sampling in the 1940 Census to create what we now call the long form; that is, we for the first time in 1940 the Census Bureau separated what as few a number of questions - i.e., the short form which went to 100 percent of the population - from a much larger number of questions on population characteristics and housing characteristics which went to a small subset of the population. And they could only do that if they had the confidence that by putting it out to only, say, one out of five or one out of six of the households they could estimate the characteristics of the entire population and its housing characteristics. So that was the first big use of sampling theory by the Census Bureau in 1940.

In the meantime, of course, you begin to get to get the emergence of a public opinion industry with George Gallup of course being one of the pioneers. Academic pioneer at this time was the National Opinion Research Center then based at the University of Chicago, which was founded in 1941 by Henry Field and his colleagues.

One of the really tough questions in democratic theory of course is to what extent is a political representative supposed to kind of use his best judgment in making laws or simply reflect the wishes and preferences of the people who put them there? This is a real key issue in democratic theory and goes all the way back to Edmund Burke, who said after all that the responsibility of the representative is not to quote/unquote "represent" constituency preferences, but to give him leadership to give them his best judgment. It even goes back to the Constitution - the whole idea of creating a republican form of government is that you would take the kind of untutored distribution of wishes and preferences and judgments of the American population, you'd filter them through this group of representatives who were after all supposed to be wiser, more intelligent, more judgmental. So this is a tough question, to what extent do you want a democracy that's based in effect upon simply the momentary preferences of the population.

Public opinion polling sort of enters this debate in a very complicated way, because it's the first time that you can rather systematically get a judgment of what a whole population feels or wants, what its priorities are, how much it weighs this versus that social policy or social good. Carried to an extreme, of course you get referendum politics. You simply put the substantive issue on the ballot, as we do in California - other states as well, but primarily in California - and then let the population vote on their judgment - on affirmative action, or immigration policy, or taxation policy, housing policy.

QUESTION: In looking at what happened in America and how we measured it in the 20th century, what is the role of Recent Social Trends?

KENNETH PREWITT: One of the really big innovations in the use of social science that has sort of tried to come to terms with the American experience was initiated in the 1920s by Herbert Hoover when he appointed a very large commission of social scientists led by William Ogburn, a great University of Chicago sociologist. And that was to produce a massive study called Recent Social Trends. And dozens and dozens of American social scientists participated in this effort, and it was to try to document everything that was going around in the population and housing characteristics as best as we could at that time. The irony was that the book was finished after Hoover had left the White House and given to Franklin Roosevelt, who didn't particularly want to be bound by some effort that a previous president had initiated. But many of the people who worked on Recent Social Trends then came into FDR's brain trust. And so a lot of the ideas in Recent Social Trends - a lot of the documentation of housing conditions, migration patterns, family structure, crime, poverty, unemployment - actually found their way into the policies in the New Deal legislation - not because the book itself was attended to, but because the people who had actually been writing it became advisors to the administration in the 1930s to work our way out of the Depression.

In some respects, though Hoover was himself a president who appreciated the importance of measuring and documenting social conditions, it really was FDR who made the most use of the social sciences. The social sciences were pretty prevalent in the First World War period, but briefly and momentarily. But not until really the FDR period do you institutionalize kind of a role for the kind of intelligence that social science can bring to the governing process. And you do that with economic advisors and social advisors. And gradually as you expand the federal responsibilities for social welfare and state of the economy, you find yourself in need of the social sciences.

QUESTION: Could you explain just briefly, what is sampling?

KENNETH PREWITT: The common-sense way to think of sampling of course is that you simply take a small part of some large population and you estimate its characteristics. Look, we take a sample of our blood to judge - make an estimate, a medical estimate about the quality of our blood, whether we are carrying diseases or not. So that's sampling theory. Or if you are doing quality control on an assembly line, you don't sort of check every Coke bottle; you check every one-hundredth Coke bottle to make sure that the recipe worked just right. So we use sampling all the time in order to kind of function. We can't sort of look at the entire population constantly; we look at a small part of that population.

Now, sampling a human population as against Coke bottles or blood is actually harder. The more heterogeneous the population that you need to make a judgment about, the harder it is to draw the perfect sample to estimate it. Blood is homogeneous, so you can take a very small sample and describe it. A human population is not homogeneous, and therefore it's a tougher task to take a good sample to estimate its characteristics.

If I simply want to estimate how many men and women there are in the American population, I can take a fairly small sample and be quite accurate in terms of the proportion of males and females. But if I want to measure its more complicated characteristics of the population - that is, not just male and female, but other demographic traits or opinions or whatever - then I have to take a somewhat larger sample.

QUESTION: Putting on your old hat as head of the Social Science Research Council, do you think that the social sciences themselves have lost a great deal of credibility because they have become so politicized?

KENNETH PREWITT: You know, if you go back to the history of the social sciences, the social sciences really become sort of independent disciplines in the 1870s, 1880s and 1890s, and that was a deliberate breakaway from politics. Up until that time, the rudimentary social sciences we had in our country at that time really were part of the social reform movement; that is, they were deliberate attempts to use information and analysis to advance one or another social cause or social reform agenda. And really the social scientists moved into a sort of scientific posture as a way to distance themselves from politics, from not being accused of having a partisan agenda. Ironically they wanted to be relevant to policy, but they thought they could be more relevant to policy - that is, the founders of the modern social sciences really in the 1910-1920 period, the Wesley Fishers for example, the social science community at that time, wanted to be more relevant to social policy by being more scientific, as they thought they could be more relevant by being less partisan. And so the origin of the kind of serious social science disciplines and journals and training programs and so forth really in the 1920s was a very self-conscious attempt to distance themselves from partisan politics in order to be useful for social policy purposes.

Now, what has interestingly happened across the century is that the social scientists in many, many ways have been brought into the policy process. After all, if you go back to the 1920s, you didn't have a large number of think tanks. You had the beginnings of the National Bureau of Economic Research early 1920s, the Social Science Research Council, the very early beginnings of the Brookings Institution - it wasn't even called that yet, but there was the beginnings of what became the Brookings Institution in that period. Well, now of course you have literally hundreds and hundreds and hundreds of social science think tanks, many of which have been established in order to advance a particular social agenda. I think what's happened and what is maybe unfortunate in this latter part of the twentieth century, first part of the twenty-first century, is that it is very difficult in Washington, D.C. to find dispassionate social science analysis. Almost all of the social science analysis that you now read about coming out of the think tanks is done because it is advancing some kind of social policy agenda, which means the earlier hope that somehow the social science community could provide reasonably independent nonpartisan intelligence that the government could then decide to use or not use has sort of been harder to maintain in the midst of a period where almost every voice, whether it's a social science voice or journalistic voice or whatever - every voice is expected to already have a predetermined political agenda.

I think that one of the things that has happened to the social science community over the last 40 or 50 years, it has become more relevant, more important. There's many, many more social policies, economic policies, other activities of our government rely upon some kind of systematic analysis, the concepts, the data, that are generated out of the social science community. Now, once your pulled that close to power it is very, very difficult not to be lined up on one side or the other.

In the university setting, where you have very strong peer pressure, you tend to get I think more dispassionate, more nonpartisan analysis than you get in the think tanks. In the think tanks - many of those have been put in place and are funded to advance a given set of social causes or political agendas. The healthy thing is that you have think tanks arguing on both sides of the issue. That is, you have more conservative and more liberal think tanks, which means they are at least again arguing about data. And that's to the good I think for a democracy.

I think there is a lot of very, very good social science. There is a lot of very good public choice theory, which is helping this country govern itself. There is a lot of good economic theory which is helping this country govern itself. The social sciences won't always get it right. It's very elusive stuff to assess what makes for a healthy economy or what makes for a functioning democracy. But I think they get us closer to having answers for those questions than in their absence. I would hate to run this country in the absence of social sciences.

QUESTION: In the course of running this monumental enterprise, the largest Census ever conducted in this country, what have you found out about America?

KENNETH PREWITT: I think the most important thing that I learned about the United States in conducting this job over the last several years would be its transformation as a society. And I am now really thinking of the growth of the new immigrant groups and how they are resettling themselves across the country. One example, I was in Columbus, Ohio, not too long ago. In 1996 it's estimated that there were about forty or fifty Somalis living in Columbus, Ohio. Today we think that number is probably fourteen or fifteen thousand.

In Charleston, South Carolina not too long ago I was told that 30 percent of the population in Charleston is now foreign-born. So it is not just that we are getting a new flow of immigrants to this society, but they are distributing themselves in unexpected places and creating whole new communities of Somalis, or the Arab Americans in, say, Dearborn, Michigan. I think that's going to change the face of this country in really a very exciting way. And these are new population groups who really want to make it as Americans - are entrepreneurial, are committed, are energized, engaged in educating their kids. They really want to kind of make their place in American society.

We now are creating a country, the first country in world history which literally has to represent all of the world, which is to say it is not just Protestant, Catholic and Jewish; it is now Hindu, Buddhist and Muslim, along with Catholic, Protestant and Jewish. It is not just four or five languages, it is 150 languages. We have the challenge and the opportunity to create the first country in world history which literally is a reflection of the entire world. If we do that well, it will be an enormous accomplishment. If we do it poorly, if we have the same kind of tensions and struggles and fights and anti-immigration and racism that we had at the turn of the last century, from roughly the 1880s to the 1920s, it will be messy, and unpleasant and nasty, and not good for this democracy, and not good for this economy.


PBS Program | Trends of the Century | Viewer's Voices | Interactivity | Teacher's Guide