look at the situation and make recommendations. You are familiar with that study? Mr. HOLMES. Yes, sir. NATIONAL ACADEMY OF SCIENCES REVIEW OF THE 1990 CENSUS Mr. MOLLOHAN. What were the recommendations coming out of the National Academy of Sciences Review of the 1990 census? Mr. HOLMES. That to continue to use the traditional census-taking methods would not improve the process; that the best method to improve that process was the use of statistical sampling, and in order to do it in the most cost-efficient fashion. Mr. MOLLOHAN. When it said it would not improve the process, did it assume an application of resources similar to those that were applied in 1990 or did it assume an increase in the application of resources? Mr. HOLMES. I do not remember the direct quote, sir, but I do vaguely remember something to the effect that no amount of money thrown at the problem would resolve it. Now, that is a paraphrase on my part. But adding additional enumerators, adding more money, would not solve the problem. Mr. MOLLOHAN. So your testimony here today is consistent with testimony you have had in the past. You have taken seriously the problems that were associated with the 1990 census, of which there was a consensus, and the recommendations coming out of the National Academy of Science study, as well as a whole bunch of expert opinion on the subject. Mr. HOLMES. Yes, sir. That is absolutely correct. Mr. MOLLOHAN. And you have come up with a plan that would incorporate those recommendations and best methodologies, which we refer to as employing sampling techniques. Mr. HOLMES. That is correct. USE OF ENUMERATORS Mr. MOLLOHAN. I would like for you to get at what I consider to be the core question or certainly one of the core questions here. This notion/premise that we can do it more accurately if we employ an actual enumeration. That seems to be at the heart of the Majority's position; that premise that if we just do more of the traditional techniques, throw more enumerators at the job, that we will get a better result than in 1990 and we will get a better result than if we were to employ sampling techniques. So I would like you to respond to that issue, discuss it, and then if you do not get at it in a way that we understand it up here, I want to talk a little bit about the Milwaukee experience that is often cited by the Majority as the good example of enumeration, which I do not think is such a good example in a final analysis. Mr. HOLMES. I guess I will start with the premise, sir, that if you continue to do things the same way you have always done them, then the outcomes are much, much more predictable. Again, having done this kind of work for 30 years and having worked on three censuses, it is clear to me that, if someone does not want to be bothered, because there are really two ways that people are missed; they are missed because we miss housing units, but more importantly Mr. MOLLOHAN. Miss individuals. Mr. HOLMES. Exactly, within household coverage. If we come to your household, no matter how many people we send there, if you decide that you are not going to tell the truth, traditional knock on the door methods will not resolve that problem. There is no way to do that. It is also important to understand that the environment in which we collect data now has changed radically over the past 20 or 30 years. As an example, the window of opportunity that you have to catch people at home now has probably dropped to maybe a couple of hours during the week and on Saturdays. I can remember 20 years ago when there was no such thing as doing an interview on Sunday. That is the most productive time to do interviews now, if you plan to catch people at home. It is clear to me, just from a layman's standpoint, this has nothing to do with statistics, but if you take a look at the direction with which voter participation has gone and responses to census forms or censuses themselves, both of them are going in the same direction, and that includes, not only that segment of the population that does not want to be counted, but your average American citizen, that, for whatever reason, they have no interest in doing it. So going back doing the same thing over and over again, no matter how much money you throw at it, does not change their behavior. Mr. PRICE. If I can just add to that a bit. The populations that had high undercounts in 1990 have grown faster. The Hispanic population, the black population, the Asian population have grown faster than the average. On that basis and the general growth of the population, we think the undercount, if we just do 1990 kind of methods, again, since the Bureau has estimated that we would have five million people undercount, instead of four million undercount in 1990 because of the two added together, the population and the relative population. It assumes that there is not a social change problem, which Jim just referred to. The American people, since 1990, continuously get more mail. They are busier. They are not at home as much. They do not want people to be asking them questions, particularly from the Government. They are less responsive, and they are more likely than 1990 not to be giving us answers or complete answers. Sampling in the ICM allows us to do more intensive data collection on a limited basis. For these 750,000 housing units, it is more expensive to send somebody to every one of those houses. If we do not do the mail, this is $3 a visit. If we do a more intensive process, we cannot afford to do that in every place in the country. But if we go there and we find out through the ICM not just the undercount, but the overcount, because people who have two residences often fill it out in Florida, and they fill it out in Michigan. They fill it out in the Ann Arbor college dorm, and they fill it out in their parents home. We have got a problem of both the overcount and the undercount that can get addressed on a sampling basis. It is expensive to do everywhere nationwide. But if we do it on a sampling basis, we can get a better overall count. Mr. MOLLOHAN. So what you are telling us is that, no matter what resources you apply to this issue, apply to the task, in tradi tional census-taking methods, putting more census enumerators on the job, you still are going to end up with an undercount that is greater and less accurate [sic] than using sampling techniques. Mr. PRICE. Yes, Congressman. People do not go around with badges saying, "I was undercounted" or "I was overcounted." Mr. MOLLOHAN. I understand. Mr. PRICE. You do not know who they are. Mr. MOLLOHAN. I understand. So what you are saying is, and the National Academy of Sciences agrees with that; is that correct? Mr. PRICE. Yes. Mr. HOLMES. That is correct. Mr. MOLLOHAN. Can you give us some sense of the professional statistical community, what their attitude is with regard to that proposition? Mr. PRICE. There is a letter signed by a number of former presidents of the American Statistical Association that endorsed the use of sampling. The American Statistical Association had a Blue Ribbon Panel on the census that has supported the use of statistical methods to improve the count. MILWAUKEE COMPLETE COUNT CAMPAIGN Mr. MOLLOHAN. Are you familiar with the Milwaukee Complete Count Campaign? Mr. PRICE. Yes. Mr. MOLLOHAN. Are you conversant with it, if I asked you some questions about it? Mr. PRICE. Yes. I, unfortunately, became an expert on it last July in response to the congressional request for a report. Mr. MOLLOHAN. Let me ask you what that demonstrated, and let me maybe lead you a little bit here. The publicity campaign that they had did achieve a better response rate than the national average, did it not? Mr. PRICE. Mail response was higher for that city than for similar cities elsewhere. Mr. MOLLOHAN. Let me finish answering [sic] the question. I mean, I know you know more about this, but just so I can hear myself and you give the answer, let me finish. So there was a better response rate to the questionnaire in Milwaukee than the national average; is that correct? Mr. PRICE. That is correct. Mr. MOLLOHAN. How much better? See, you just were not that conversant with it. [Laughter.] Mr. PRICE. I wrote it here in the report we gave. Mr. MOLLOHAN. Would it be accurate, to refresh your memory here, if Milwaukee's response rate was 76 percent? Mr. PRICE. And the national average is 74 percent. Mr. MOLLOHAN. And the national average was 74. So with their big publicity effort, they got a 2 percent increase in the response rate; is that correct? The national average response rate; is that correct? Mr. PRICE. They were above average, but not much. Mr. MOLLOHAN. They were above average, which means, if it is average, somebody else was above average, too. Mr. PRICE. Right. Mr. MOLLOHAN. And they did not, arguably, at least they do not have the reputation for putting on this kind of a program. And that Milwaukee, with its tremendous campaign, ranked third, not first, in mail return rates; is that correct? Mr. PRICE. That is correct. Mr. MOLLOHAN. And it still left 24 percent of the occupied units in the city to be counted by door-to-door enumerators; is that correct? Mr. PRICE. Yes. Mr. MOLLOHAN. So, at the end of all of that effort did they put more enumerators on the job? I know they used civic organizations and all of that. What did they do with regard to enumerators? Mr. PRICE. The unique thing about Milwaukee was what the city and the State put resources into to trying to raise public awareness of the census and to make forms available in the "Were You Counted?" campaign later on. But the Bureau's efforts, in terms of following up on those 24 percent, were comparable, the enumerators to follow-up those 24 percent were comparable to other places with 24 percent of the housing units that did not reply by mail. Mr. MOLLOHAN. At the end of the day, the post-enumeration survey indicated that approximately 2.3 percent of the city's residents were missed; is that correct? Mr. PRICE. Yes. Mr. MOLLOHAN. Was that higher or lower than the national average? Mr. PRICE. The national average was 1.6 percent. Mr. MOLLOHAN. Is not the point here that there is just a ceiling that cannot really be raised by traditional methods that you bump up against here in terms of achieving accuracy? Perhaps it is 2.3 percent of the population you are going to miss or perhaps it is 2.1 or 1.9, but everything we know about it that through traditional enumeration methods you are going to be inaccurate and have an undercount of approximately that percentage. Am I Mr. PRICE. Yes. That is Census Bureau's view. That is the view of the statistical experts on this. Mr. MOLLOHAN. So that is really the bottom line, is it not? No matter how much money we spend on additional enumerators and publicity advertising, questionnaires, we are still going to have those people, that percentage of people that do not want to be counted or just fall through the cracks some way, and the only way to get them is through these sampling techniques to get a more accurate result. Mr. PRICE. And we used those methods in the agricultural census. We used exactly the same kinds of methods. The way it is proposed to be done by the Census Bureau is now done by the agricultural statistics operation, but they do a sampling for people who do not respond by mail and they do a sampling to correct for those who had systematic undercounting, and they get better results because they use statistics. The question really is are we going to be banned from using sta COST OF CENSUS COMPARED TO 1990 CENSUS Mr. MOLLOHAN. And there is one more question, too. How much is it going to cost to increase these enumerators and in the end result achieve, a less accurate result? Do you have any estimates? Mr. PRICE. We estimated in a different context before the Fiscal Year 1998 and Fiscal Year 1999 dual track funds came along. Last spring we gave estimates, if you recall, $675 million to $800 million extra above $4.0 billion to do a design very similar to 1990 that excluded sampling-over and above the sampling design that we had in mind in the prior plan. Mr. MOLLOHAN. Okay. Could you talk about the attempt to do the 2000 Census through enumeration methods in terms of how much it would cost. We have already I hope in my line of questioning at least established that. Mr. PRICE. No. Our full cycle cost is on the order of $4 billion with the sampling plan. It would cost on the order of $600 to $800 million more to do the 1990 style but that still leaves you with 5 million people undercounted. And the cost of trying to reduce that 5 million-because you do not know who is undercounted, you just know that there is an undercount―to try to go and find the needles in the haystacks, to go find the houses with the tenant for which they do not have everybody actually living in that residence on their completed form, is very expensive. Mr. MOLLOHAN. All right. Can you give us an estimate of how expensive? Mr. PRICE. We will be working on trying to do some kinds of estimates but it is very hard to do because we do not Mr. MOLLOHAN. So, it is your testimony that if you were to do a 1990 census it would cost you $500, $600, $700 million more in 2000 than it would a sampling census? Mr. PRICE. That is correct. Mr. MOLLOHAN. Just off the top, to do what you did in 1990? Mr. MOLLOHAN. And then to try and improve the accuracy of that count by applying additional enumerators you are coming up with a number and you do not have a number of what that might be? Mr. PRICE. NO. And it would depend on what techniques were used. We think that it is-we have not spent a lot of time investigating that up until November because we thought it was so exorbitantly expensive for so little an amount of return. But since November, we are operating under a different mandate so we are supposed to get as close to 100 percent as possible, using non-sampling techniques, and we are prepared to do that and we are doing studies to try and do that. It will show very expensive costs for very little additional return to accuracy but we are now starting to work on that. Mr. MOLLOHAN. Is not a way to say that, an exorbitant cost for a less accurate census than sampling? Mr. PRICE. Yes. Mr. MOLLOHAN. An exorbitant cost for a less accurate census. |