• Home
• Light Cigarettes
• Studies & Briefs
• Expert Testimony
• Quitting Resources
• Contact Us

1 THE COURT: You may be seated and good

2 afternoon. Any motions?

3 MR. WAGNER: (Shook head).

4 THE COURT: Call your next witness, please.

5 MR. SWEDLOW: Michael Dennis.

6 THE COURT: Sir, stand up and raise you are

7 right hand to be sworn.

8 MICHAEL DENNIS

9 the Witness herein, called as a witness on

10 Plaintiff's behalf, being first duly sworn, was

11 examined and testified as follows:

12 DIRECT EXAMINATION

13 QUESTIONS BY MR. SWEDLOW:

14 Q. Good afternoon. Please state your name and

15 address for the record?

16 A. My name is Michael Dennis. I live at 1118

17 Forrestal, Foster City, California.

18 Q. Where are you currently employed?

19 A. At a company called Knowledge Networks.

20 Q. What is your current job title at

21 Knowledge Networks?

22 A. My job title is Vice President and

23 Managing Director of the Government and Academic

24 Research Department.








3


1 Q. We'll come back to your job title at your

2 company Knowledge Networks later. Let's back up for

3 a second. Where did you good to college and when

4 did you graduate?

5 A. I went to the University of Texas

6 Graduating with a Bachelor in Government in 1984,

7 followed that year up with a Masters in Government

8 at the University of Texas.

9 Q. After you finished your Masters in

10 Government at the University of Texas did you go on

11 to get further education?

12 A. Yes. I went to the University of Chicago

13 finishing my Doctorate in Political Science.

14 Q. What was the focus of your doctoral

15 studies?

16 A. I focused on biomedical research,

17 biomedical politics, quantitative studies and other

18 similar researches.

19 Q. Turn back to where you work currently

20 today, what is the business of Knowledge Networks?

21 A. Knowledge Networks is a full service

22 market research and social science research firm. We

23 conduct surveys on the Internet. We have a special

24 core called a web-enabled panel. We provide full








4


1 scale assistance for all our customers from

2 questionnaire design, research design, all the way

3 through data collection and analysis.

4 Q. When you say for all your consumers, who

5 are your particular customers or clients?

6 A. My customers tend to be university

7 professors and other academics who have won grants

8 or contracts from the Federal government for

9 Federally sponsored research.

10 Q. This Federally sponsored research, is this

11 Internet based survey research?

12 A. That's correct.

13 Q. How long have you personally been involved

14 as a survey research practitioner?

15 A. Approximately 13 years.

16 Q. And how much of that time has your primary

17 job been to collect and analyze survey data?

18 A. For the last 11 years.

19 Q. Turning back now to what you do at

20 Knowledge Networks. What are your primary job

21 responsibilities?

22 A. My day to day responsibilities consist of

23 managing the government academic department. This is

24 the client side part of the work as opposed to








5


1 operations. I work daily with my university based

2 customers and Federal agencies to design surveys,

3 implement the survey, collect the data, work within

4 the company to improve all our data collection and

5 quality control procedures and ultimately deliver

6 data and reports to our customers.

7 Q. Where did you work before you worked at

8 Knowledge Networks?

9 A. I worked as a social sciences consulting

10 firm called Abt Associates.

11 Q. Abt Associates. What were your job

12 responsibilities at Abt Association?

13 A. Over the course of the years I was there I

14 was the Survey Director managing data collection

15 projects for a number of Federally sponsored

16 projects. Later on I was a Senior Scientist and

17 Associate Project Director of very some very large,

18 fairly responsive projects.

19 Q. Is it fair to say your work Abt Associates

20 was primarily devoted to survey research and

21 analysis?

22 A. That's correct.

23 Q. In your 13 year career as a survey

24 research practitioner, approximately how many








6


1 surveys, how many substantive surveys have you

2 either drafted yourself or collaborated in the

3 drafting process?

4 A. Over the time a minimum of 40 perhaps 50

5 questionnaires. Many of them in the last few years

6 working with my academic customers at Knowledge

7 Networks. Many of them also for setting our internal

8 data base that we are required to set up at

9 Knowledge Networks so we can understand the

10 demographic, behavioral, attitudinal and health

11 related characteristics of your research panel

12 members.

13 Q. When you say these internal surveys, what

14 information are you trying to capture with these

15 internal surveys you are talking about with

16 Knowledge Networks?

17 A. There were approximately fifteen to twenty

18 of these surveys and the topics are very wide

19 ranging from health related issues such as people's

20 disease and chronic illness states and histories, to

21 questionnaires about durable good consumption, what

22 cars you might own, what kind of cars, to politics,

23 social activism, civic beliefs. It's a wide range of

24 surveys and topics.








7


1 Q. In your career have you ever done any work

2 as a survey research practitioner for the Center for

3 Disease Control or CDC?

4 A. When I was the Abt Associates I was the

5 associate practitioner scientist for the National

6 Immunization Survey.

7 Q. What is the, in relative size, what is the

8 National Immunization Survey?

9 A. The NIS, National Immunization Survey is

10 the largest Federally Sponsored random digit dialing

11 telephone survey project that is sponsored by the

12 Federal government.

13 Q. Can you define what random digit dialing

14 is?

15 A. Random digit dialing telephone surveys is

16 a widely practiced methodology for collecting

17 attitudinal data over the phone. It consists of

18 randomly selecting telephone numbers, whether

19 directory listed or not for initiating a contact to

20 a household and ultimately collecting data from that

21 household.

22 Q. And in your work as a survey research

23 practitioner have you done any work for the Social

24 Security Administration?








8


1 A. Yes. Again when I was Abt Associates I was

2 project director for the data collection for a study

3 called the Project Network Follow-up. That was a

4 survey where we did a follow-up survey of the Social

5 Security disability beneficiaries who had completed

6 a survey three years previously. The topic of that

7 particular project involved factors that could

8 determine which social security disability

9 beneficiaries are able to return to work versus

10 those that remained disabled and not able to return

11 to work.

12 Q. Have you done any surveys or work for the

13 National Drug Control Policy?

14 A. Yes. Sometimes called the White House drug

15 tzar office, ONDCP, in '95 '96, again I was at Abt

16 Associates. I was project director of a study called

17 Estimating the Prevalence of Hard Core Drug Use in

18 Cook County, Illinois. That was a relatively

19 complicated data collection project where first

20 stage of recruitment would be institutions that

21 housed our research subjects. So we needed to

22 gaining the cooperation of the booking facilities of

23 Cook County Jail, many drug treatment centers and

24 homeless shelters throughout the Chicago area; and








9


1 the secondarily the potential hard core drug users

2 that we surveyed.

3 Q. Have you ever done any work for the

4 National Cancer Institute or what we have been

5 calling the NCI?

6 A. Yes. For NCI when I was at Abt Associates

7 I was the project director for a study called a Case

8 Control Study of Stomach Cancer in Polish Americans.

9 And that was a study funded by NCI. Our principle

10 investigator came from the University of Illinois

11 Chicago. Our job there was to recruit individuals

12 who have stomach cancer for intensive surveys about

13 the diets and past practices as well as a core of

14 individuals who did not have stomach cancer.

15 Q. What was the size of the samples for this

16 survey?

17 A. That was a relatively small sample survey.

18 In total there was about 200 individuals with

19 stomach cancer that was part of the project and a

20 corresponding number of control or people who did

21 not have stomach cancer.

22 Q. Have you during your career as a survey

23 researcher have you written any publications?

24 A. I've written several publications.








10


1 Q. Can you identify let's call it the

2 highlights of your publication career?

3 A. I think the highlights would focus on peer

4 reviewed articles. In the September 11th, 2002 issue

5 of the Journal of the American Medical Association I

6 contributed some information that lead to a peer

7 review article there. I was not coauthor on that

8 particular article, that was authored by Professor

9 Roxanne Silver of the University of California,

10 Irvine. However, a month previously in the Journal

11 of the American Medical Association I was coauthor

12 for an article with Bill Schlenger from the Research

13 Triangle Institute. We think that's a landmark

14 article for estimating the effects of the 911

15 terrorist attacks on literally for New Yorkers,

16 people in the Washington D.C. area, as well as

17 across the entire country. In 1998 I was coauthor of

18 an article that was published in Health Services

19 Research, again peer reviewed article that examined

20 factors that determined which patients are put on

21 kidney transplant waiting lists versus those

22 patients that are not placed on kidney transplant

23 waiting lists. Which harkens back to my dissertation

24 research. Finally there are a number of articles








11


1 which relate to survey methodological issues that

2 have been published in the proceedings of the

3 American Statistical Association and other places.

4 Q. What is that, the proceedings of the

5 American Statistical Association?

6 A. The proceedings of the American

7 Statistical Association is mostly a compilation of

8 papers and presentations that have been made at the

9 Annual American Statistical Association meeting and

10 also some of the other conferences that are

11 associated with the ASA.

12 Q. What is the Association for Public Opinion

13 Research?

14 A. AAPOR, A-A-P-O-R, American Association of

15 Public Opinion Research is made up of Survey

16 Research Practitioners like myself. It also consists

17 of individuals who aren't necessarily practitioners

18 of surveys, but are survey research methodologists,

19 statisticians who may have teaching and academic

20 responsibilities and then finally political

21 pollsters are frequently in the AAPOR annual

22 meeting.

23 Q. We flew pass the description of Knowledge

24 Networks survey web -- web-enabled panel








12


1 methodology. I want to spend little more time

2 discussing that. Can you describe this web-panel

3 approach it survey implementation?

4 A. Yes. The web-enabled panel is different

5 from other web-enabled panels. In the sense that it

6 is a random probability sample of United States

7 households. As a random sample of U.S. households it

8 consists of households that both have Internet

9 access prior to our contacting them and households

10 that never had Internet access before. What is

11 distinguishable about the Knowledge Networks

12 approach is that it is a random sample

13 representative cross section of the United States;

14 and, secondly we do provide the Internet appliances

15 and Internet service so that these randomly selected

16 households can participate in our research surveys.

17 Q. How many randomly selected households do

18 you currently have in the Knowledge Networks panel?

19 A. Approximately 40,000.

20 Q. And we were discussing earlier surveys

21 that you conducted that you called internal surveys,

22 are those the panel members for which these surveys

23 are applied to?

24 A. That's correct.








13


1 Q. What information is gathered with those

2 surveys?

3 A. These internal survey, which we call

4 profile surveys, covers a wide range of areas,

5 health, politics, social attitudes and beliefs,

6 practices, home ownership status. Literally about

7 2000 characteristics at this point have been

8 collected on most of our panel members.

9 Q. Is the Knowledge Networks web-enabled

10 panel designed to track the United States population

11 in terms of demographic characteristics?

12 A. It is.

13 Q. Let me ask -- why don't you define what a

14 demographic characteristic is?

15 A. A demographic characteristic would be for

16 example the age of a respondent or gender, household

17 income, education status and other like

18 characteristics.

19 Q. How does the Knowledge Networks panel

20 compare with respect to the demographic information

21 about the general United States population?

22 A. It tracks very closely to the United

23 States population characteristics as documented by

24 the United States census.








14


1 Q. Has Knowledge Networks won any awards from

2 you called it AAPOR, the survey organization in the

3 recent years?

4 A. That's correct. Approximately -- it

5 wasn't the last meeting, but the year before we won

6 the innovators award which was given to the two

7 founders of our company, Norman Nie and Professor

8 Rivers from Stanford University.

9 Q. Why -- let me ask you this, what is so

10 innovative about the Knowledge Networks' approach to

11 survey methodology?

12 A. What is innovative is that prior to

13 Professor Rivers and Nie no one thought Internet

14 surveys could be done with a probability sample, and

15 secondly with a national representative sample. The

16 assumption was that Internet base surveys could only

17 be conducted with households that already had

18 Internet access. The breakthrough for Knowledge

19 Networks was to have a random sample of all

20 households regardless what their prior Internet

21 status was.

22 Q. If somebody agrees to be in your panel you

23 give them the equipment they would need to get on

24 the Internet?








15


1 A. That's correct.

2 Q. What do you give them?

3 A. Currently we give them a Web TV Internet

4 appliance.

5 Q. Do you pay them for their time in

6 completing these surveys?

7 A. No, we don't.

8 Q. So what do they give you then in response

9 to receiving this Internet access?

10 A. In exchange we ask for their participation

11 in our attitudinal and other surveys.

12 Q. Is their participation in this panel

13 contingent upon answering any particular survey?

14 A. No, it's not contingent on that.

15 Q. Is the participation in that panel

16 contingent upon answering any particular question in

17 any particular?

18 A. They can skip a question or refuse to

19 answer any question.

20 Q. Do these panel members have what is called

21 a Bill of Rights?

22 A. Yes. There's a Bill of Rights on a piece

23 of paper as well as posted on the Website that lists

24 all of their rights as a Knowledge Networks panel








16


1 member. Explaining their rights for privacy,

2 confidentiality of their information and ultimately

3 the voluntary nature of their participation in the

4 panel.

5 Q. How many surveys approximately do you ask

6 these panels to complete each month or week?

7 A. On average it's about three a month. The

8 absolute maximum is four.

9 Q. How long on average are the surveys that

10 you ask them to complete?

11 A. On average a survey is about ten to

12 fifteen minutes.

13 Q. How in your opinion does the length and

14 complexity of this survey compare to the other

15 surveys that you delivered to your panel members?

16 A. I consider the survey conducted in this

17 case to be shorter and less complex than our average

18 survey.

19 Q. Going back to Knowledge Networks panel

20 generally and survey delivery devices that you use.

21 Are there recognized advantages to the Knowledge

22 Networks survey method as compared to other survey

23 methods?

24 A. Yes. There are various advantages that my








17


1 consumers who tend to be university based survey

2 researcher, principle investigators, mentioned to

3 us. Some of them involve the self-administrative

4 form of data collection.

5 Q. What do you mean self-administrative?

6 A. The self-administered surveys are

7 distinguished by the fact there is no interviewer

8 involved whether on the telephone or in face to face

9 type situation. Our respondents sit in front of

10 their TVs and with a keyboard or remote control

11 administer the surveys to themselves.

12 Q. Why is that an advantage to having the

13 survey be self-administered?

14 A. There are some drawbacks of having

15 interviewers involved. Survey research interviewers

16 can sometimes have their own subjective judgments.

17 But more commonly there is an interaction that can

18 occur between a respondent and the interviewer that

19 can cause respondents to provide biasing responses

20 and in our field that's called socially desirable

21 responses where the respondent is answering in a way

22 to please the interviewer.

23 Q. That does not exist in your survey

24 delivery mechanism with no interviewer?








18


1 A. That's correct.

2 Q. What other advantages exists or recognized

3 for this delivery mechanism?

4 A. Many of our customers, again university

5 based academics and the Federal agencies that we

6 deal with prefer our methodology because of the

7 visual aspect of the survey taking experience.

8 Because the survey questions are displayed on a

9 screen as opposed to being read orally over the

10 telephone or in a face to face interviewer situation

11 the respondent can see the entire question on a

12 single screen and take their time and read and even

13 reread that question to make sure that respondent

14 really understands the question. In a telephone base

15 survey environment that kind of rereading and

16 comprehension of questions in our opinion and the

17 opinion of many of my customers are compromised.

18 Q. When you say they can take their time, do

19 these respondents in this survey that we are talking

20 about today or any of your surveys, do they have any

21 limitations at all on the amount of time they can

22 spend answering a particular question?

23 A. No. No limitation at all.

24 Q. Do they have limitations on the amount of








19


1 time they can take to answer the survey itself?

2 A. No limitation.

3 Q. Do they have the requirement to complete

4 the survey in one sitting?

5 A. No. They can interrupt and come back later

6 that day or a different day.

7 Q. Is there any limitation as to the time of

8 day or the day in which they can take this survey?

9 A. No. They can take the survey any time of

10 day or night. In fact our research shows that much

11 of the survey taking occurs after 9:00 at night.

12 Q. Are there any other, I know I'm testing

13 your memory, are there any other recognized

14 advantages that you can think of right now with

15 respect to the Knowledge Networks survey

16 methodology?

17 A. There are several things, I'll mention

18 two. One has to do with the efficient targeting of

19 subject populations. Because we have this profile

20 data base previously collected information of panel

21 members demographic and characteristics, we are able

22 to field a survey and assign it, if you will, to

23 specific supopulations. For example diabetics, a

24 pharmaceutical comes to me as asks me for a sample








20


1 of diabetics. We've already identified that sample,

2 so it's possible to field the survey quickly.

3 The second other advantage I would mention

4 is rapid data delivery because it is an Internet

5 base form data collection it's possible that we can

6 deliver survey data and results in a few days or a

7 few weeks.

8 Q. I want to run through some organizations

9 and ask you if they have been a client of Knowledge

10 Networks. Don't give me a long description of the

11 project for two reasons, one if it's confidential I

12 don't want to violate confidentiality; and, two it's

13 Friday afternoon. First the National Science

14 Foundation, are they your client?

15 A. The National Science Foundation sponsors

16 various of my customers with grants.

17 Q. How about the Environmental Protection

18 Agency?

19 A. Similarly they've award grants or

20 contracts to some of my consumers who subsequently

21 subcontracted to me for data collection.

22 Q. I want to explain that process you were

23 talking about. When I say was the Environmental

24 Protection Agency a client of yours, and you said








21


1 they subcontracted. What does that mean?

2 A. That means I do not work, my company does

3 not work directly for the EPA, there is another firm

4 that serves as the prime contracting organization

5 and then I'm a subcontractor to the prime

6 contracting organization.

7 Q. And are you currently working in a

8 subcontractor capacity for any prime contractor

9 getting a grant from the EPA?

10 A. Yes.

11 Q. Who is that prime contractor that you have

12 a relationship with?

13 A. There are several.

14 Q. Are those confidential relationships?

15 A. I believe they are a matter of public

16 record.

17 Q. Who are those people?

18 A. There's -- I'll mention a few that come to

19 mind. There's Professor J.R. DeShazo at UCLA.

20 Professor Kip Fiscusi from Harvard Law. There's an

21 individual named Bob Rowe from Stratus Consulting.

22 Those are the immediate customers that come to mind.

23 Q. Now we understand the relationship between

24 the Federal agencies and your firm, does your firm








22


1 do any work for the Veterans Affairs Department?

2 A. Yes.

3 Q. How about for the CDC, I know you said you

4 did some work for them at Abt, does your current

5 firm do work for the CDC?

6 A. That's correct. Again through a

7 subcontracting relationship.

8 Q. I want to ask you about a few foundations

9 and ask if they are also clients from your current

10 company. The Hewlitt Foundation?

11 A. Yes.

12 Q. The Robert Woods Johnson Foundation?

13 A. Yes.

14 Q. And the Sloan Foundation?

15 A. Yes.

16 Q. I don't think that I have, I'm not going

17 to identify all of your clients here today, but

18 we've been focusing on the academic and government

19 clients. Do you have any news media type clients?

20 A. We do work occasionally for CBS news.

21 Q. Have you done a project recently for CBS

22 news?

23 A. Yes. We covered the very recent State of

24 the Union address for CBS news.








23


1 Q. Describe that project a little bit?

2 A. We provided the information that went on

3 air approximately ten minutes after the Democratic

4 response was finished. We gave an instant read, if

5 you will, on the American public opinion in terms of

6 their reactions to President Bush's State of the

7 Union address as well as the Democratic response.

8 Q. How were you able to capture the instant

9 reaction of the public to the State of the Union

10 address?

11 A. We pre-recruited a sample of viewers

12 before the State of the Union Address and asked them

13 to watch the State of the Union Address and the

14 Democratic response while simultaneously answering

15 questions and then giving us their second by second

16 feedback on their favorable versus unfavorable

17 ratings of the President and Democratic responses.

18 Q. Based upon all of your experience in

19 survey research both at Knowledge Networks and Abt

20 Associates and your education, does the Knowledge

21 Networks web-enabled panel survey methodology reduce

22 what you refer to as social desirability of a survey

23 deliverer?

24 A. Yes.








24


1 Q. Does this Knowledge Networks methodology

2 increase the comprehension of questions for the

3 respondents?

4 A. I believe it does because of the visual

5 aspect of the way the questionnaire is presented to

6 the research subjects.

7 Q. Now I want to switch to the survey you

8 conducted for the purpose of this case. I believe

9 your survey is already offered into evidence. I'll

10 hand you a copy of what is Group Exhibit 74a. I

11 don't need you to identify and authenticate it

12 because this is already into evidence. Let's jump to

13 the description of this survey. As I understand this

14 survey has three separate parts, is that correct?

15 A. That's correct.

16 Q. Can you describe the first, the very first

17 part, what the purpose of the very first part of the

18 survey was?

19 A. The first part of the survey, I refer to

20 is the screening portion of the survey. The

21 screening questions. And the purpose of the

22 screening questions is to identify the respondents

23 that should be interviewed for the study, i.e.

24 Marlboro Lights smokers.








25


1 Q. Briefly identify the purpose of what we'll

2 call the second part of the survey?

3 A. The second part of the survey, which is

4 really the first part of the main interview,

5 consists of questions intended to measure what the

6 brand descriptors of Marlboro Lights mean to our

7 research subject, the Marlboro Lights smoker.

8 Q. And now please briefly describe the third

9 part of the survey what the purpose of that part is?

10 A. The third part of the survey again, not to

11 be confusing, that's the part two of the main

12 interview, the second half of the main interview, is

13 intended to measure the value of the harm reduction

14 benefit of the Marlboro Lights.

15 Q. If we can display the first slide. I want

16 to discuss the way that this survey was conducted

17 from the beginning to end not necessarily with the

18 questions, but with the way people wound up in the

19 survey. Please display the first slide. How does

20 somebody become what you characterize here as an

21 invited respondents?

22 A. First of all, they are all Knowledge

23 Networks panel members, currently active on the

24 panel. Secondly, these are all respondents from whom








26


1 we have collected information previously on their

2 smoking status. We have information on file about

3 whether they're current smokers or recent smokers.

4 We selected randomly from among the group that

5 previously reported that they were either current

6 smokers or recent smokers. We didn't have brand

7 specific information on file of what cigarettes they

8 smoked we just know they were recent smokers or

9 current smokers.

10 Q. First of all, did you create this

11 presentation we are about to see here?

12 A. I drew it up on a piece of paper and

13 designed it, yes.

14 Q. How does a person get invited as a

15 respondent?

16 A. The person gets invited by sending an

17 e-mail to them. That e-mail has the information in

18 it inviting them to participate in the survey.

19 Q. Pop up the next one. Then here you

20 identify completed screening survey. What does that

21 mean?

22 A. Of the 2,701 panel members to whom we sent

23 the invitation, 1,779 completed the first few

24 questions, the screening portion of the survey, for








27


1 a completion rate of about 66 percent.

2 Q. We'll take the screen portion in a minute

3 let's finish out the slide. Then of the, I think

4 this becomes clear when we take the survey, of the

5 people who completed the screening survey, then you

6 have qualified for main interview, explain that?

7 A. The screening questions in the first part

8 of the survey identify the eligible or qualified

9 respondents. There were a total of 276 persons out

10 of the base of 1,779 that qualified for the main

11 interview by virtue of their responses of the first

12 three questions.

13 Q. Do you know how this 15.5 percent Marlboro

14 Lights percentage smoker compares to the Marlboro

15 Lights smokers in the general smoking population?

16 A. It is my understanding there is another

17 data source from the year 2001 that reports a number

18 that's less than a percentage point off that one.

19 Q. Here it says completed part one of main

20 interview: Meanings of brand Descriptors, it has

21 100 percent. What does that mean?

22 A. That means that all persons that were

23 qualified for the part one of the main interview

24 actually completed part one of the main interview.








28


1 Q. What does that mean?

2 A. Similarly. All persons that were eligible

3 for part two completed part two.

4 Q. Let's take the screening portion of the

5 survey. Can you go to the next slide? Now, is this

6 how the survey appears exactly to the person who

7 sits at their TV taking the survey?

8 A. It is exactly the same with very minor

9 formatting differences because this screen capture

10 was from a personal computer instead of from a Web

11 TV.

12 Q. I'll read this question. "Now please

13 select the primary cigarette brand that you have

14 smoked in the past year." In fact I think we would

15 like to go back one slide to -- this is the first

16 question. "Before we get started with this important

17 study about what cigarette names and descriptors

18 mean to you, please select the primary cigarette

19 brand that you are currently smoking." And here --

20 who chose these brands to display here?

21 A. I selected those brands.

22 Q. Why did you select so many brands?

23 A. I thought it was important to camouflage

24 or disguise as much as possible which brand would








29


1 need to be selected to continue to the main part of

2 the interview. So I put together a longer list of

3 brands to disguise which possible brand it might be

4 that is the main one for the study.

5 Q. Why were you trying to disguise the brand

6 that was the subject matter of the study?

7 A. In my experience it's possible that

8 research subjects could potentially try to screen

9 into a survey or screen out of a survey by trying to

10 divine what the questionnaire designer is trying to

11 do.

12 Q. Let's say we clicked on not currently

13 smoking, can we go to the next screen? Is this the

14 question you would get if you clicked on not

15 currently smoking?

16 A. That's correct.

17 Q. What's the purpose of this next question?

18 A. If the person is not currently smoking

19 then we have a follow-up question that asks which

20 primary cigarette brand they may have smoked in the

21 past year.

22 Q. "Now please select the primary cigarette

23 brand that you have smoked in the past year." Here

24 we have the same brands or different brands?








30


1 A. The same brands.

2 Q. And what if you clicked have not smoked in

3 the past year, what happens to you?

4 A. If you clicked that then the survey is

5 over for the research subject.

6 Q. What if you click anything other than

7 Marlboro on this screen?

8 A. The survey is over.

9 Q. What if you clicked any other brand other

10 than Marlboro on the previous screen?

11 A. The same applies. You have to select

12 Marlboro on the first screen or second screen.

13 Q. Okay. Now this was a screen we went to if

14 we chose not currently smoking. What happens if you

15 chose Marlboro on the previous screen?

16 A. Then you go to the next screen. That

17 follow-up with a question with what kind of Marlboro

18 cigarette.

19 Q. This question, "Please selected the type

20 of Marlboro cigarette that you have smoked on a

21 regular basis in the past year." And you provide

22 choices. Who qualifies out of these choices for the

23 survey?

24 A. Only those persons that selected the








31


1 Marlboro Lights or Marlboro Lights 100's.

2 Q. When we were going through these three

3 questions some people got sent to a second question,

4 some people skipped the second question and went to

5 the third question. How does that happen and what is

6 that called?

7 A. That's called a skip pattern in the

8 questionnaire.

9 Q. What's the purpose of skip patterns?

10 A. Skip patterns are very common,

11 particularly in computer-assisted interviewing, for

12 assuring research subjects answer only those

13 questions that are appropriate for them.

14 Q. Why did you limit the eligibility of this

15 survey to people who are currently smoking or people

16 who have smoked within the past year?

17 A. The main reason I would cite is that the

18 first part of the questionnaire, I should say the

19 part one of the main interview asks questions about

20 people's initial reasons for smoking Marlboro

21 Lights. For that reason it is important to select

22 persons that are current or recent smokers because

23 these are the individuals that are very much

24 recently engaged in smoking practices and therefore








32


1 would have a much higher ability to recall their

2 initial reasons for smoking Marlboro Lights.

3 Q. Did the substance of part two figure into

4 your decision to only include current smokers and

5 recent smokers as well?

6 A. Yes. That is the case because the part

7 two is situated in the present tense for

8 understanding why people and how much they value the

9 two cigarette products that are offered in part two.

10 Q. Now, we just discussed skip patterns,

11 let's say somebody chooses Marlboro Lights or

12 Marlboro Lights 100's so they qualify for the

13 survey. What's the next thing that they would see?

14 A. They see a transition screen. That reads

15 "Now we are going to ask you a series of questions

16 about your decision to smoke Marlboro Lights" --

17 Q. I'll finish the sentence -- "a lowered tar

18 and nicotine cigarette - as opposed to a regular

19 cigarette." What kind of screen is this?

20 A. A transition screen.

21 Q. What is the purpose?

22 A. It's very common in survey research to

23 introduce a new concept sometimes in the Internet

24 motive data collection with a separate screen that's








33


1 a bridge between one section of the questionnaire to

2 a different section of a questionnaire.

3 Q. Let's click, continue, you would get the

4 next question, right?

5 A. Yes.

6 Q. Before we run through this whole survey,

7 we were previously discussing skip patterns. Is the

8 presentation we are seeing today the same

9 presentation that one individual respondent would

10 get?

11 A. It is. We are going to be going through

12 different paths of the questionnaire in the

13 presentation, so therefore if someone were to select

14 no, for this question, they would not see the

15 question we are about to show on this because there

16 is a skip pattern.

17 Q. Where I make a mistake about who goes

18 where, please correct me. If I am saying the wrong

19 people are going to be asked this question explain

20 that to the Court. The first question, "Can you

21 recall why you initially chose to smoke a light or

22 lowered tar and nicotine cigarette on a regular

23 basis as opposed to a regular cigarette." Now what

24 happens to the people who select yes here?








34


1 A. The people who select yes here are

2 administered the rest of the questions from part one

3 of the main interview.

4 Q. What happens to the people who select no

5 here?

6 A. The people that respond no are not

7 administered the questions from part one of the main

8 interview.

9 Q. Why didn't you ask the people who said no

10 here the first part of the survey?

11 A. The part one questions, with the exception

12 of this one, are substantive questions about

13 people's attitudes towards the brand descriptors and

14 those attitudes are referring back to the initial

15 reasons why people smoke the Marlboro Lights brand.

16 It was important to identify those research subjects

17 that can remember their initial reasons why they

18 smoked Marlboro Lights. Because all the part one

19 questions refer back to that period in time.

20 Q. Here we have the results. Can you read

21 those results for us?

22 A. Sure. The results show that about 77

23 percent of the respondents could remember their

24 initial reasons why they smoked Marlboro Lights and








35


1 about 23 percent could not.

2 Q. And the significance of this question, if

3 I understand it, if you said yes you get asked the

4 first part of the survey, if you said no you moved

5 to the second part of the survey?

6 A. Yes.

7 Q. Let's move to the first part of the

8 survey. These are the people who said yes. The

9 question is, "Think back to when you started

10 regularly smoking Marlboro Lights - did you think

11 then that the words 'lowered tar and nicotine' that

12 appeared on the cigarette package meant that a

13 Marlboro Lights cigarette would deliver less tar to

14 your body than a regular cigarette like Marlboro

15 Reds." Then let's pretend we chose the answer and

16 see what the results were. What are the results of

17 that question?

18 A. The results show that the vast majority of

19 the respondents when they thought back to that time

20 when they started smoking Marlboro Lights regularly

21 thought that the words lower tar and nicotine, 83

22 percent, meant that the Marlboro Lights cigarettes

23 would deliver less tar and only about 17 percent

24 thought it would not.








36


1 Q. Now if you said yes to this question where

2 did you go?

3 A. There's a follow-up question, question

4 three.

5 Q. Question three says, "Did your belief that

6 a Marlboro Lights cigarette would deliver less tar

7 to your body contribute to your decision to smoke

8 Marlboro Lights as opposed to smoking a regular

9 cigarette like Marlboro Reds." Let's pretend we

10 selected something. And what are the results of this

11 question?

12 A. You will see that again the vast majority

13 of the respondents thought that their belief

14 contributed to their decision to smoke the Marlboro

15 Lights. It's about 87 percent.

16 Q. Next question please? This question, "Did

17 you think then that the words 'lower tar and

18 nicotine' that appeared on the cigarette package

19 meant that a Marlboro Lights cigarette was less

20 harmful or safer than a regular cigarette like

21 Marlboro Reds." These were asked of the same people

22 who got asked if lower tar and nicotine meant lower

23 tar, less tar to your body, is that correct?

24 A. This question was administered to all








37


1 persons who could remember their initial reason to

2 smoke Marlboro Lights.

3 Q. Let's click on an answer. These results?

4 A. These results show that almost 70 percent,

5 68 percent thought that the words lower tar and

6 nicotine meant that a Marlboro Lights cigarette was

7 less harmful or safer.

8 Q. Of these people who said yes, let's go to

9 their follow-up question. They are asked, "Did your

10 belief that Marlboro Lights were less harmful or

11 safer contribute to your decision to smoke Marlboro

12 Lights as opposed to a regular cigarette like

13 Marlboro Reds." What are the results here?

14 A. The results show virtually everybody, in

15 this case 96 percent, thought their belief that the

16 cigarette was less harmful or safer contributed to

17 their decision to actually smoke that cigarette.

18 Q. These are the people that said yes to the

19 previous question and skipped to this question?

20 A. That's correct.

21 Q. Let's go to the next question. Well, this

22 is not a question. This is -- what are these called

23 again?

24 A. Transition screen.








38


1 Q. Before we go to this transition screen.

2 There are other questions in this survey, aren't

3 there, that aren't being displayed now from this

4 first series, is that correct?

5 A. Yes, they are.

6 Q. What is the difference in form between the

7 questions that aren't being displayed and questions

8 that's are being displayed?

9 A. We are displaying the responses to the

10 close-ended questions. The questions that we are

11 not showing right now are the open-ended questions.

12 Q. What's the difference between a closed and

13 open-ended questions?

14 A. A close-ended question such as ones we are

15 seeing with a yes and no response categories provide

16 a finite list of response categories for the

17 respondent to actually selected.

18 Q. Why did you choose close-ended questions

19 for these, why did you choose that form for these

20 questions here?

21 A. The close-ended form has several

22 advantages, I'll mention two. The first one is that

23 it removes the subjectivity element biasing

24 possibility. If you have an open-ended question








39


1 where respondents are providing answers in their own

2 words decisions have to be made, often subjective

3 decisions, about how to actually code those

4 responses into real categories. I wanted to avoid

5 that situation where an element of subjectivity was

6 brought in to actually code all the answers that

7 would be written out in real words.

8 The second reason is that the close-ended

9 form allows us to program and skip logic so that if

10 someone says yes to a particular question then there

11 is an appropriate follow-up question. With an

12 open-ended type question where you have literally

13 just a lot of words from the respondent as they are

14 answered it's much more difficult to program a skip

15 pattern.

16 Q. Doctor, turn back to this transition

17 screen, is that right?

18 A. Yes.

19 Q. It says, "We have finished the series of

20 questions related to the words 'lower tar and

21 nicotine.' Now we are going to ask you a series of

22 questions about the word 'Lights' that also appears

23 on the package of Marlboro Lights." The first

24 question in that series, "Did you think then that








40


1 the word 'Lights' in the name meant that a Marlboro

2 Lights cigarette would deliver less tar to your body

3 than a regular cigarette like Marlboro Reds?" And

4 before we show the results, this question looks a

5 lot like a prior set. What's the difference about

6 this set than the prior set?

7 A. The only difference is the term lights.

8 The new brand descriptive has been introduced,

9 lights. First few questions in part one of the main

10 interview were about a different brand descriptor,

11 lower tar and nicotine.

12 Q. And the results of this?

13 A. Again it's about 77 percent of the

14 individuals thought that the word Lights in this

15 case, not lower tar and nicotine, but Lights meant

16 to them that a Marlboro Lights cigarette delivered

17 less tar.

18 Q. And for people who said yes here they were

19 asked the following question?

20 A. Yes. They were asked the follow-up

21 question on whether that contributed to their belief

22 or not.

23 Q. And this question says, "Did your belief

24 that Marlboro Lights cigarette would deliver less








41


1 tar to your body contribute to your decision to

2 smoke Marlboro Lights as opposed to a regular

3 cigarette like Marlboro Reds." And the results of

4 this?

5 A. Similar results. Still above 90 percent.

6 At this point 93 percent of those individuals

7 thought their belief contributed to their decision

8 to smoke that cigarette.

9 Q. Now moving through to the survey

10 question. The survey is already a matter of

11 record. I'm going to have you paraphrase the

12 question being asked here.

13 A. Sure. Did you think the words Lights in

14 the name meant Marlboro Lights was less harmful or

15 safer than a regular cigarette like Marlboro Reds.

16 Q. And the results of this question?

17 A. It's still above 70 percent thought that

18 the word Lights meant in this case less harmful or

19 safer.

20 Q. And turning to the next question. "Did

21 your belief that Marlboro Lights were less harmful

22 or safer contribute to your decision to smoke

23 Marlboro Lights as opposed to a regular cigarette

24 like Marlboro Reds." And the results of this?








42


1 A. It's about 96 percent thought their belief

2 that Marlboro Lights were less harmful or safer

3 contributed to their decision. A very small

4 percentage said no.

5 Q. Next question please? Let's take a step

6 back. This now, we are moving into a new section of

7 the survey, is that correct?

8 A. That's correct.

9 Q. What are you aiming at in this new section

10 of the survey?

11 A. We are now in the part two of the main

12 interview. And what we are intending to do here is

13 measure the value of the harm reduction attribute of

14 Marlboro Lights.

15 Q. When you say measure the value of the harm

16 reduction attribute. How did you do that in this

17 survey?

18 A. The first step is set up a comparison and

19 questionnaire between the two cigarette products.

20 Q. And what are those two cigarette products?

21 A. The first one is the Marlboro Lights

22 cigarette that delivers less tar and less harmful or

23 safer than Marlboro Reds. The second cigarette

24 product is exactly the same except it is just as








43


1 harmful and delivers the same amount of tar as

2 Marlboro Reds. Just to summarize, there is less

3 harmful or safer cigarette product versus just as

4 harmful cigarette product. In both cases the

5 reference is the Marlboro Reds cigarettes.

6 Q. That these two products we are comparing,

7 if I understand you correctly, a less harmful light

8 to a just as harmful light, is that correct?

9 A. Yes, that's correct.

10 Q. Outside of the context of surveys, can you

11 explain this valuation of an attribute concept using

12 an example?

13 A. Yes. I think an example most people could

14 relate to would be freshness and baked goods or

15 bakery products. One can think of the example of

16 bagels or doughnuts in terms of how fresh they are.

17 And to the extent to which people value freshness in

18 a bagel when they go to a store and purchase one.

19 Q. What would be the example you would set up

20 if you wanted to determine how people value

21 freshness in bagels?

22 A. I would set up two products for

23 comparison. I would set up a product that people are

24 used to seeing, a bagel that's made on a same day








44


1 basis versus a bagel that is less fresh, for example

2 a bagel that had been made the day before.

3 Q. So you would compare, if a bagel store

4 owner wanted you to determine what discount he would

5 have to provide for a day old bagels, you would set

6 up a survey or some measurement method to compare a

7 fresh bagel that he sells to a day old bagel, is

8 that correct?

9 A. That's correct.

10 Q. How would you measure the value of that,

11 the difference between a fresh bagel and a day old

12 bagel?

13 A. You would look at what people are willing

14 to spend for the product that doesn't have this

15 freshness attribute. What are people willing to

16 spend, or conversely what kind of price discount

17 would you have to give the research subject in order

18 to buy the day old bagel.

19 Q. Let's go read this question. "Suppose you

20 had a choice between two different kinds of Marlboro

21 Lights. Both taste exactly the same and cost the

22 same and are identical in every way to the Marlboro

23 Lights you smoke, except that these two cigarettes

24 differ in the amount of tar they deliver to your








45


1 body and how harmful they are." And this first is

2 what we already discussed setting up the comparison,

3 is that correct?

4 A. That's correct.

5 Q. "One delivers less tar and is less harmful

6 or safer than Marlboro Reds. The other delivers the

7 same amount of tar as Marlboro Reds and just as

8 harmful as Marlboro Reds. Which one would you buy?"

9 And here we have two choices. Let's stay here for a

10 second. You repeat the choices in the answer as

11 opposed to just leaving them in the question, why

12 did you do that?

13 A. That was important to present a stark

14 visual contrast between the two alternatives and put

15 them in a similar place on the screen so the

16 respondents can clearly see the differences between

17 the two products.

18 Q. Is this an example of why the visual

19 presentation is more understandable than just the

20 audio over the telephone?

21 A. Yes, it is. In a telephone form of this

22 survey the respondent most likely keeps upper most

23 in their minds the last read response option.

24 Q. Here our choices we have, "The Marlboro








46


1 Lights cigarette that delivers less tar and is less

2 harmful or safer than Marlboro Reds." And, "the

3 Marlboro Lights cigarette that delivers the same

4 amount of tar as Marlboro Reds and is just as

5 harmful as a Marlboro Reds." And let's see the

6 results. And you are not -- here not surprisingly

7 almost 93 percent choose the less harmful or safer

8 alternative, is that correct?

9 A. That's correct.

10 Q. Moving to the next slide. After that first

11 question, do you have -- if we were back at the

12 bagel store, do you have the difference in value

13 between a day old bagel and the fresh bagel?

14 A. No, you don't have that valuation of the

15 difference yet.

16 Q. For purposes of this question what is

17 being compared here? What's changed about this

18 question?

19 A. This question is the first attempt to try

20 to get a point estimate, I should say the first step

21 to get a point estimate of the extent to which the

22 respondents value the harm reduction attribute of

23 Marlboro Lights. In this case we are asking people

24 to compare the less harmful cigarette that is at the








47


1 same price or usual price they pay for Marlboro

2 Lights and compare that to a just as harmful

3 Marlboro Lights that is at half price. Se we are

4 asking them if they are willing to buy the just as

5 harmful cigarette if given a 50 percent discount.

6 Q. We are comparing a less harmful Lights at

7 full price to a just as harmful Lights at half

8 price?

9 A. That's correct.

10 Q. Let's -- "Suppose you could buy a Marlboro

11 Lights that delivers the same amount of tar as a

12 Marlboro Reds and just as harmful as Marlboro Reds

13 for half the price you usually pay for Marlboro

14 Lights. If the following cigarette tasted exactly

15 the same, which would you buy?" And here your two

16 choices: "The Marlboro Lights that deliver less tar

17 and less harmful or safer than Marlboro Reds at the

18 same price you usually pay for Marlboro Lights."

19 That's your first choice. Less harmful price you

20 pay. Or, "The Marlboro Lights cigarette that

21 delivers the same amount of tar as the Marlboro Reds

22 and is just as harmful as the Marlboro Reds at half

23 the price of Marlboro Lights." Then let's see the

24 results. What are these results?








48


1 A. The results show that people by in large

2 prefer the less harmful or safer cigarette at the

3 full price and relatively few are willing to buy the

4 just as harmful cigarette at the half price.

5 Q. We have 84.9 percent preferring the full

6 price less harmful, and 15.1 who wants the half

7 price just as harmful. Now are we done? Do we have

8 a point estimate for these people what they value

9 the health attribute?

10 A. We don't have a point estimate. We know

11 that for fifteen percent of these particular

12 respondents that they value the just as harmful at a

13 50 percent rate, but we don't have a point estimate.

14 It could still be higher than 50 percent. We need

15 to ask a follow-up question later to get a follow-up

16 question later.

17 Q. Let's follow the 85 percent. For the next

18 question. What's the purpose of this question?

19 A. The point of this question is to ask if

20 there's any lower price at all that these

21 respondents would be willing to purchase the just as

22 harmful cigarette.

23 Q. "Is there any lower price at which you

24 would buy Marlboro Lights that delivers the same








49


1 amount of tar as Marlboro Reds and is just as

2 harmful as the Marlboro Reds if a Marlboro Lights

3 that delivers less tar and is less harmful or safer

4 than a Marlboro Reds was available at the price you

5 usually pay for Marlboro Lights? Remember that both

6 Marlboro Lights cigarettes taste exactly the same."

7 So just focusing on the first sentence, the question

8 is, is there any lower price in which you would buy

9 a Marlboro Lights that delivers the same amount of

10 tar and just as harmful, is that correct?

11 A. Yes.

12 Q. What were the results of this?

13 A. The results were most people were

14 unwilling to buy just as harmful cigarettes at any

15 price. It was 72 percent reported that there was not

16 a lower price that could entice them to purchase the

17 just as harmful cigarette.

18 Q. So for these, this is 72.2 percent is just

19 the people asked this question. Do you know what the

20 total number of people who actually said no?

21 A. Yes. 156 people.

22 Q. Out of a total survey size of what?

23 A. 276 representing about 66 percent of the

24 entire survey sample.








50


1 Q. Over half of the -- do we now have a point

2 estimate for their valuation?

3 A. We know they don't value the -- they have

4 a high value for the harm reduction attribute.

5 Q. Now, if we were back in the bagel store

6 what bagel consumer is this?

7 A. This would be the consumer that would not

8 purchase a day old bagel at any price.

9 Q. He would go in and buy a fresh bagel for a

10 dollar or whatever the price?

11 A. Whatever the price would be this person

12 would pay it.

13 Q. Now, for the people who, the 156 people,

14 are they done with the survey?

15 A. They are done with the survey.

16 Q. How many valuation questions did these

17 people have to answer?

18 A. A total of three questions.

19 Q. Let's back up now. I realize we are going

20 to ask and answer more questions than the survey

21 respondents. I want to explain to the Court the way

22 this process works. Let's pretend that back in the

23 question where we asked do you want a full price

24 less harmful or half price just as harmful, that








51


1 they chose the half price just as harmful, what

2 happens to those people?

3 A. Those people receive a follow-up question,

4 I believe it's on the next screen, that asks about

5 their point estimate, about how much they value that

6 just as harmful cigarette.

7 Q. If I understand correctly, you know they

8 will pay half price, now you want to know what the

9 actual price is?

10 A. Yes. Are they willing to spend even more

11 than 50 percent of the usual price for a Marlboro

12 Lights in exchange for that just as harmful

13 cigarette.

14 Q. "What is the highest price you would pay

15 for a Marlboro Lights that delivers the same amount

16 of tar as the Marlboro Reds and is just as harmful

17 as Marlboro Reds if a Marlboro Lights that delivers

18 less tar and is less harmful or safer than a

19 Marlboro Reds was available at the price you usually

20 pay for Marlboro Lights?" Here we're asking them

21 what's the highest price they are willing to pay for

22 the just as harmful light, is that correct?

23 A. Yes.

24 Q. Here it's listed from 90 to 50, did








52


1 everybody get it in that list?

2 A. Half the sample received it in that order

3 from 50 to 90, and half the sample received the

4 order from 50 to 90. So half in descending order and

5 half in ascending order.

6 Q. Why did you radomize or change the order

7 of these numbers?

8 A. It's a common practice to reduce the

9 possibility of any bias or error that can be created

10 if people do not read the entire list. There's a

11 tendency, I think a very small one, but a tendency

12 nonetheless for a respondent sometimes not to read

13 the entire list, but select the one they think

14 matches their best response.

15 Q. Let's see the results of this question.

16 Can you describe what happened here with these

17 results?

18 A. You will see all the options were selected

19 at least once. Most people, I should say the

20 sporality (phonetic) response was 70 percent of the

21 price you usually pay for Marlboro Lights.

22 Q. So at this point we know for these people

23 the discount they would need to buy the cigarette

24 without the reduced harm?








53


1 A. Yes. Now we have a point estimate for

2 this particular group.

3 Q. Some people would be willing to buy it at

4 a 30 percent discount or at 70 percent of the

5 price. So we have a point estimate discount for all

6 of these people, is that correct?

7 A. That's correct.

8 Q. Let's go to the next slide. Now the first

9 comparison was between a what we call a less harmful

10 cigarette and a just as harmful cigarette. Did you

11 do another comparison between two different kinds of

12 cigarettes?

13 A. Yes, we did. In this second half of the

14 part two of the main interview we set up a

15 comparison of the less harmful or safer cigarette,

16 which we just talked about, versus a new concept and

17 product which is the could be more harmful than

18 Marlboro Reds cigarettes.

19 Q. Now we are comparing a less harmful

20 cigarette to a could be more harmful cigarette?

21 A. Yes.

22 Q. Both referring to what?

23 A. Always referring to the Marlboro Reds.

24 Q. Taking us back to the bagel store, what is








54


1 the equivalent here of what we are comparing?

2 A. I think the equivalence in that particular

3 analogy would be to a bagel that's even more stale

4 than a one day old bagel. A bagel that may have been

5 made two days before. So the question would be, do

6 want to buy a bagel that's fresh made on the same

7 day versus a bagel that could be more stale.

8 Q. So the freshness attribute in that example

9 is even more pronounced, is that correct?

10 A. That's correct.

11 Q. In this example the harm attribute is, the

12 difference is more pronounced, is that correct?

13 A. That's correct.

14 Q. Let's read this question. "Suppose there

15 was another kind of Marlboro Lights that could be

16 more harmful than Marlboro Reds because it had both

17 an increased potential for causing genetic and

18 chromosomal damage and contained higher levels of

19 some of the most toxic substances found in cigarette

20 smoke as compared to a Marlboro Reds cigarette.

21 Which of the following cigarettes would you buy?

22 Remembering that both Marlboro Lights cigarettes

23 taste exactly the same." Now, we know why these

24 words appear in this survey from our prospective,








55


1 tell me why these words appear in your survey from

2 your prospective.

3 A. From my prospective this particular

4 question has new information about additional health

5 risks that is the best information available from

6 the science in terms of additional health risks from

7 smoking Marlboro Lights.

8 Q. Maybe we should take a step back here. Did

9 you draft this survey?

10 A. Yes.

11 Q. Did the lawyers draft this survey?

12 A. No.

13 Q. Did the lawyers draft any of these

14 questions?

15 A. No.

16 Q. Where did you get specifically these words

17 this increased potential for causing genetic and

18 chromosomal damage and higher levels of some of the

19 most toxic substances found in cigarette smoke?

20 A. There are requirements given to me for

21 concepts that needed to be measured for this project

22 from plaintiff's counsel.

23 Q. Now here the choices are a less harmful

24 cigarette and a could be more harmful cigarette. In








56


1 the selection choices we have: "Marlboro Lights

2 cigarette that delivers less tar and is less harmful

3 or safer than Marlboro Reds at the price you usually

4 pay for Marlboro Lights." That's the full price less

5 harmful. Compared -- you can choose "the Marlboro

6 Lights cigarette that could be more harmful than

7 Marlboro Reds at half the price you usually pay for

8 Marlboro Lights." Now, who is asked these questions

9 these could be more harmful set of questions?

10 A. The only people asked these questions are

11 people that place some value on the just as harmful

12 Lights that we discussed earlier.

13 Q. These are people who are willing to buy

14 the day old bagel so to speak at some discount?

15 A. That's right.

16 Q. Now we are asking them if they want the

17 two day old bagel at some discount?

18 A. That's correct.

19 Q. Let's see the results of this. What are

20 results of this?

21 A. By in large, less than a quarter were

22 willing to buy the half price or could be more

23 harmful. This is the two day old bagel analogy. 76

24 percent were willing to pay for the full price for








57


1 the less harmful light.

2 Q. I should have asked you to define this

3 earlier. This valid percent term that appears here.

4 That term refers only to the people asked this

5 question, is that correct?

6 A. That's correct.

7 Q. So if you weren't asked the question you

8 are not counted in this percent?

9 A. That's correct. It also includes or should

10 say excludes any cases where the respondent didn't

11 answer the question. And some of the questions we

12 had one respondent that would not answer the

13 question. So it's not exactly just those not asked

14 the question, there's a very small number situation

15 where people refused to answer the question.

16 Q. What was the none response rate generally

17 for this survey question by question?

18 A. One divided by 276.

19 Q. One divided by 276?

20 A. A quarter of a percent.

21 Q. Let's say that we followed the people who

22 -- let's go to the next slide, I can't remember who

23 we followed. These are the people who said that they

24 wouldn't buy -- let me ask this. Does this second








58


1 set of valuation questions track exactly the first

2 set of valuation questions?

3 A. It's the same structure for writing out a

4 point estimate that we had in the first half of the

5 part two.

6 Q. This question is asking the same question

7 we had before, is there any lower price, is that

8 correct?

9 A. That's correct.

10 Q. Let's see the results of this survey.

11 Now, here we have, what's the result here?

12 A. Most respondents would not buy the could

13 be more harmful cigarette even if given a very high

14 discount. There's really no lower price that these

15 respondents would be enticed or attracted to buy the

16 could be more harmful or two day bagel analogy, they

17 are not willing to do that.

18 Q. And then for the 21 percent or so, are

19 they done with the survey?

20 A. No, they are not done. There's a follow-up

21 question for them.

22 Q. Let's see that follow-up question. Here

23 they are being asked the highest price they would

24 pay. They are spread between ten and forty percent,








59


1 why is that?

2 A. Because there's a separate parallel path

3 for those who are willing to pay half price. That

4 will get us a point estimate between 50 percent and

5 90 percent. For those for whom we know their

6 valuation is less than 50 percent, they are given

7 this follow-up question which logically is trying to

8 derive a point estimate between the 10 and 40

9 percent of the usual price paid for the Marlboro

10 Lights.

11 Q. Now at this point even though we took more

12 questions than the average survey taker, as I

13 understand it -- let's go back and show the

14 results. Here are the results of that question.

15 This shows a spread among divided these options as

16 well. For these people, we have a point estimate, is

17 that correct?

18 A. That's correct.

19 Q. Now moving to the last question. Was this

20 last question that you constructed used as you

21 understand it in the damages model for this case?

22 A. No, it was not used in my understanding in

23 the damages model.

24 Q. Do you know what the purpose of this








60


1 question was?

2 A. My understanding is the purpose is to get

3 an understanding or measure the impact of an

4 addiction on willingness to pay for the could be

5 more harmful cigarette.

6 Q. Who was asked this last question?

7 A. Only those people that could place or did

8 place some value on the could be more harmful

9 cigarette.


24 Q. I'll read the question into the record.








61


1 "Now suppose you are not addicted to cigarettes in

2 any way. Is there any price at which you would buy

3 the kind of Marlboro Lights that could be more

4 harmful than Marlboro Reds because it has both an

5 increased potential for causing genetic and

6 chromosomal damage and contained higher levels of

7 some of the most toxic substances found in cigarette

8 smoke as compared to the smoke from a Marlboro

9 Reds." What these -- this question was only asked of

10 those who wanted the equivalent of two day old

11 bagel, is that correct?

12 A. That's correct.

13 Q. At some discount?

14 A. Yes.

15 Q. What were the results of this survey?

16 A. The results show that, it's about a 50/50

17 split. About a 51 percent split. So some people

18 that even if they were not addicted could be

19 possibly tempted to purchase this could be more

20 harmful cigarette if given a discount.

21 Q. Now, everyone is done with the survey?

22 A. That's correct.

23 Q. I asked you this before, but I guess I'll

24 ask you this again. How does this survey compare to








62


1 other surveys delivered by your company to the panel

2 members?

3 A. I think this particular survey

4 questionnaire is relatively simple for comprehension

5 particularly for surveys of this type.

6 Q. How does this survey compare, when you say

7 surveys of this type, how does this survey compare

8 in relative difficulty to comparative valuation

9 surveys generally?

10 A. Generally, I would put this on the easiest

11 or low end of that spectrum of relatively easy to

12 difficult cognitively challenging questionnaires.

13 This particular questionnaire involves a research

14 subject that is deeply familiar and personal to the

15 research subject, their smoking. A lot of the

16 comparative type questionnaires that I assist in

17 designing and ultimately implementing involve much

18 more remote topics. Things that don't necessarily

19 have an impact on their daily lives or they think

20 about on a daily basis.

21 Q. Was there a pretest performed for this

22 survey?

23 A. Yes.

24 Q. What is the purpose generally of a








63


1 pretest?

2 A. The purpose of a pretest is to check all

3 phases of the questionnaire. To make sure that

4 respondents understand the questions. To give

5 ourselves an opportunity to perform quality control

6 on the program version of the questionnaire itself.

7 Ultimately it's a full system check, if you will, on

8 all phases of the project including even the

9 sampling of the project.

10 Q. Now you are under oath, Dr. Dennis, and

11 I'm going to ask you this question, were any of the

12 changes made from the pretest to the main interview

13 made to influence the direction of the results?

14 A. Absolutely not.

15 Q. Does this survey accurately measure the

16 respondent's reported beliefs and valuations?

17 A. Yes, I believe so.

18 Q. Do you believe based upon your experience

19 as a survey research practitioner that this survey

20 produces reliable results?

21 A. I believe this survey, if conducted again

22 in a matter of days would produce the same results.

23 Q. Now, I think you just defined for me in

24 your answer, what does reliable mean?








64


1 A. Reliable in this context would mean

2 reproducibility of the survey holding all other

3 things constant such as information flows and

4 attitudinal changes in society.

5 Q. In your opinion as a survey research

6 practitioner are these survey questions

7 comprehensible to the respondents?

8 A. Yes. In my opinion they are.

9 Q. Were the response options clear?

10 A. The response options were clear. I think I

11 was very careful in designing and contrasting

12 products and making this appear on the screen so

13 they could be easily read.

14 Q. Was the survey administered objectively?

15 A. The surveys were self-administered.

16 There's no possibility for interviewer lead bias, so

17 the answer is yes.

18 Q. In your expert opinion as a survey

19 research practitioner do these results accurately

20 measure the reported attitudes and valuations of

21 these respondents?

22 A. I believe so. It measures their reported

23 attitudes.

24 Q. In your expert opinion as a survey








65


1 research practitioner, do these results accurately

2 measure these characteristics in the population of

3 Illinois Marlboro Lights smoker?


6 (At this point a brief recess

7 was taken.)



18 THE COURT: You may resume.

19 Q. (By Mr. Swedlow) I'm going to back up one

20 question just so the record is clear. In your expert

21 opinion as a survey research practitioner do your

22 survey results in this case accurately measure the

23 reported attitudes and valuations of the

24 respondents?








73


1 A. Yes.

2 Q. In your expert opinion as a survey

3 research practitioner and -- excuse me, in your

4 expert opinion as a survey research practitioner do

5 your survey results accurately measure the reported

6 attitudes and valuations in the population of the

7 United States Marlboro Lights smokers?

8 A. Yes.

9 Q. And in your expert opinion as a survey

10 research practitioner do your survey results

11 accurately measure the reported attitudes and

12 valuations in the population of Illinois Marlboro

13 Lights smoker?

14 MR. WAGNER: Object, your Honor. There is

15 no foundation for him to testify to that.

16 THE COURT: Overruled.

17 A. I believe the results are fully

18 projectable to the Illinois residents Marlboro

19 Lights smoker.

20 Q. We just had a foundation objection, you

21 have no idea what that means. I'll ask you a

22 question related that. On what do you base that

23 opinion, that your survey results being nationally

24 representative of Marlboro Lights smokers is also








74


1 representative of Illinois Marlboro Lights smokers?

2 A. My opinion is based on my understanding of

3 the demographic characteristics of Illinois

4 residence as being broadly representative of the

5 United States as a whole. I see no reason why

6 estimates for the national population of Marlboro

7 Lights smokers would not be projectable and relevant

8 for the Illinois residence of Marlboro Lights

9 smoker.

10 MR. SWEDLOW: I have no further questions.

11 I want to offer into evidence the demonstratives of

12 the demonstrative exhibits.

13 THE COURT: I was wondering about that.

14 MR. SWEDLOW: It's Group Exhibit Number

15 98A.

16 I will note for the record the direct

17 examination was one hour and three minutes.

18 THE COURT: You will have one hour and I'll

19 give you an hour and twenty minutes.

20 CROSS EXAMINATION

21 QUESTIONS BY MR. WAGNER:

22 MR. WAGNER: Thank you, your Honor.

23 Q. Dr. Dennis, we haven't met before. My name

24 is Jeff Wagner I represent Philip Morris. Welcome to








75


1 town. Dr. Dennis, I'm going to discuss your survey

2 in as much detail as I can get into in an hour and

3 twenty minutes. Let's get a few items established

4 off the top. Okay. If I say KN that's Knowledge

5 Network. Is that okay with you?

6 A. Yes.

7 Q. This is not a survey of the class, is it?

8 A. I don't have the working definition of

9 class to answer your question.

10 Q. The class involves all persons who

11 purchased Marlboro Lights and Cambridge Lights in

12 Illinois from 1971 to 2001, you did not purport to

13 do a survey of the class, did you?

14 A. I designed a survey to measure as close as

15 possible the members of the class.

16 Q. Did you in fact survey class members, did

17 you reach out and survey and ask people whether or

18 not they purchased Marlboro Lights or Cambridge

19 Lights in Illinois from 1971 to 2001?

20 A. That question was not.

21 Q. So you don't know whether this surveyed

22 the population of the class, do you?

23 A. I know it surveyed a fraction of the

24 class, I can't say it surveyed the Illinois








76


1 residents as you just defined them.

2 Q. Okay. The survey is not limited to

3 Illinois residents, is it?

4 A. It's a national survey.

5 Q. Is it limited to Illinois residents?

6 A. No.

7 Q. Thank you.



21 Q. Dr. Dennis, the class period as I said

22 once is from '71 to February 2000, your survey has

23 several willingness to pay questions, questions 12

24 to 17, right?








77


1 A. Correct.

2 Q. Those questions were asked in October or

3 November of 2002, correct?

4 A. Correct.

5 Q. Those questions asked about what a person

6 would do now in 2002, I mean not now today, but in

7 2002 in a hypothetical situation. They don't ask

8 about the past, do they?

9 A. No, they don't.

10 Q. Okay. If I say WTP, I mean willingness to

11 pay?

12 A. Yes.

13 Q. The survey is not about Cambridge Lights

14 at all?

15 A. There is not a reference to Cambridge

16 Lights.

17 Q. Now, let's talk about your survey results

18 what they mean, okay? What do they mean?

19 Reliability is a word that has a particular meaning

20 in field survey research, correct?

21 A. Correct.

22 Q. Reliability means the ability to collect

23 the same information twice using the same questions,

24 correct?








78


1 A. Correct.

2 Q. Okay. So reliability means replicability,

3 would you agree with me on that?

4 A. Yes, I would.

5 Q. Okay. Reliability though is different than

6 validity, correct?

7 A. Correct.

8 Q. Okay. And a survey can be reliable but

9 still produce bias or invalid results?

10 A. That's correct.

11 Q. And when survey researchers talk about

12 producing accurate answers they use the term

13 validity, correct?

14 A. That's correct.

15 Q. Now, validity refers to whether the survey

16 answers corresponds to a true population measure,

17 correct?

18 A. Correct.

19 Q. And reliability and -- validity is

20 typically unknowable, isn't it?

21 A. Validity typically is unknowable.

22 Q. So what experts in your field do is they

23 try to find the most comprehensive data set that we

24 can use as a benchmark to see if your survey results








79


1 match up?

2 A. That's one method, but not the only one.

3 Q. You have no such external data set which

4 you can measure the results from your survey

5 against, right?

6 A. That's correct. It's not an unusual

7 situation, that's correct.

8 Q. You are not offering an opinion here that

9 this survey provides valid estimates concerning what

10 people believe or their attitudes toward Marlboro

11 Lights cigarettes, correct?

12 A. I'm offering an opinion if asked about the

13 validity of self-reported attitudes.

14 Q. You have not made any assessment of the

15 extent to which these self-reported measures are the

16 true population measures, correct?

17 A. I have not used an external or comparative

18 data set to make that valuation. However, I've made

19 my own valuation based on observations of the

20 process.

21 Q. My question is whether or not you have

22 made an assessment of the extent to which these

23 self-reported measures are the true population

24 measures?








80


1 MR. SWEDLOW: I object. That was just

2 answered.

3 THE COURT: I thought you asked and I

4 thought he answered it.

5 MR. WAGNER: I thought he did not give me a

6 clear answer. And so I would like to make sure I

7 have a clear answer on the record as to whether or

8 not this witness has been.

9 THE COURT: You know, I'll let you ask it

10 one more time.

11 Q. Have you made an assessment -- isn't is

12 true, Dr. Dennis, that you intend to report the

13 results of this survey as reflecting the

14 self-reported attitudes of this targeted

15 subpopulation, your survey respondents, without

16 making an assessment of the extent to which these

17 self-reported measures are the true population

18 measures?

19 A. To be very clear. I had no intentions

20 because it's not feasible to make an external

21 validity check against a comparison benchmark and

22 data set for the reason that that benchmark and data

23 set is unavailable. However, based on my survey

24 research experience and based on my intense personal








81


1 familiarity with the way the survey was done I am

2 capable of making my own professional assessment

3 about the validity of this research process.

4 Q. Understood. Let me ask the question this

5 way then. It is your opinion, is it not, that this

6 survey is unable to measure what the true population

7 of Marlboro Lights smokers think objectively about

8 the brand, correct?

9 A. By objectively if you mean the true

10 beliefs that people have as opposed to counter

11 distinction about the self-reported attitudes I

12 agree with your statement.

13 Q. I want to jump to the second part of the

14 survey, second part of the interview that you talk

15 about which are the damages questions, the

16 willingness to pay questions, okay? Now, questions

17 12 through 18 are those willingness to pay and

18 damages questions, correct?

19 A. Excuse me, 12 through 17.

20 Q. And those questions ask people

21 hypothetical questions about their willingness to

22 pay for Marlboro Lights or a hypothetical cigarette,

23 right?

24 A. For products as defined.








82


1 Q. And the product that was defined, you

2 understand you were given scientific information by

3 the lawyers about Marlboro Lights. They told you

4 that it was not any safer at delivering lower tar

5 and nicotine than a regular Marlboro?

6 A. I lost you in the middle part of your

7 sentence. Could you repeat it?

8 Q. When you were framing the question you

9 were told by plaintiff's counsel, by Mr. Swedlow,

10 that Marlboro Lights do not deliver any lower tar

11 and nicotine and therefore are just as harmful as

12 Marlboro Reds, correct?

13 A. The information provided for question

14 fifteen and sixteen regarding the chromosomal damage

15 and toxic substances, that's what you are referring,

16 that's information provided to me as the scientific

17 evidence which Mr. Swedlow provided.

18 Q. Then you were hypothesizing about a

19 Marlboro Lights that could be safer than Marlboro

20 Reds, correct?

21 A. That could be more safer, correct.

22 Q. Okay. So none of these questions actually

23 ask respondents what they would be willing to pay

24 for the hypothetical cigarette, do they?








83


1 MR. SWEDLOW: I'm going to object. I

2 don't understand the question.

3 THE COURT: I don't care, I understand it.

4 Overruled.

5 A. Could you please repeat the question?

6 Q. Sure. You were -- I want to know whether

7 or not your questions actually asked respondents

8 what they would be willing to pay for the could be

9 safer hypothetical cigarette?

10 A. That's correct.

11 Q. Let's see if we can get some additional

12 information here about the willingness to pay

13 questions. Does the survey ever ask respondents what

14 price people are willing to pay for Marlboro Lights?

15 A. It asks them what you are willing to pay

16 for the just as harmful or could be harmful

17 cigarettes.

18 Q. Let me clarify, does the survey ever ask

19 them in dollars and cents what is the price you are

20 willing to pay for Marlboro Lights?

21 A. It does not ask the question in terms of

22 dollars and cents.

23 Q. Does the survey ask them what price in

24 dollars and cents they are willing to pay for








84


1 Marlboro Lights if they know it's not less harmful

2 than a Marlboro Reds?

3 A. I think I answered that question. There is

4 no responses in here correlating to price in terms

5 of dollars and cents.

6 Q. You have to bear with me. How about this,

7 does the survey ask respondents what price in

8 dollars and cents they are willing to pay for

9 Marlboro Lights if its more harmful or could be more

10 harmful than Marlboro Reds?

11 A. The answer is no.

12 Q. Try one more. Does the survey ask

13 respondents what price in dollars and cents they are

14 willing to pay for Marlboro Lights if a really safer

15 Marlboro Lights exists?

16 A. No.

17 Q. What your survey asks is the willingness

18 it pay in relative terms, a willingness to pay

19 relative to another product that might or might not

20 be on the market, right?

21 A. That's correct.

22 Q. Your survey doesn't tell us for example

23 whether people would be willing to pay ten dollars

24 for a pack of safer Marlboro Lights cigarettes,








85


1 right?

2 A. No.

3 Q. People might be willing to pay fifteen

4 dollars a pack for safer Marlboro Lights, your

5 survey doesn't tell us that?

6 A. No.

7 Q. It could be that people in your survey are

8 willing to pay more than the current market price

9 for a pack of a Marlboro Lights cigarette even

10 knowing that it's not safer, your survey doesn't ask

11 them in dollar or cents what they are willing to

12 pay?

13 A. It does not ask dollar and cents about

14 their willingness to pay.

15 Q. In these questions, questions 12 through

16 17, I think we said these questions ask people what

17 they are going to do in November of 2002 at the time

18 of the survey, it doesn't ask about the past?

19 A. Correct.

20 Q. These questions don't ask respondents what

21 they would be willing to pay for these different

22 cigarettes at any time in the past?

23 A. Reality, all in present. Present tense.

24 Q. If we are asking about today all these








86


1 answers remain true, right?

2 A. That's correct.

3 Q. No for each of these. If we are asking

4 about the past, no, no, no, no, right?

5 A. That's correct.

6 Q. And even if you are asking about relative

7 willingness to pay between two products, this survey

8 doesn't ask them about their relative willingness to

9 pay for the past, does it?

10 A. No, it doesn't.

11 Q. All right. The questions that you asked

12 relating to willingness to pay and damages, Dr.

13 Dennis, those are the type of questions commonly

14 referred to as contingent valuation questions,

15 right?

16 A. Correct.

17 Q. And you understand that, do you not, that

18 a panel of the United States government has issued

19 guidelines for contingent valuation surveys how they

20 ought to be designed, right?

21 A. Yes, I'm aware of that.

22 Q. These are the NOAA guidelines from the

23 National Oceanic and Atmospheric Administration?

24 A. Correct.








87


1 Q. I'm going to hand you Exhibit 4348.

2 MR. WAGNER: I only have two copies, your

3 Honor.

4 THE COURT: Don't worry about it. Take care

5 of the witness. Why don't you look at this one.

6 MR. WAGNER: I have mine. There are only

7 two extras.

8 Q. If we turn to page fifteen -- let's first

9 talk about what the NOAA guidelines are. You said

10 you are familiar with these.

11 MR. WAGNER: If we can put up 4348.2.

12 Q. If you could highlight the middle

13 paragraph that begins, NOAA has published several

14 federation notices. I want to make sure the Court

15 has an understanding of what the NOAA panel is and

16 what they published. They went through seven federal

17 register notices requesting information and comments

18 on approaches to developing damage assessment

19 procedures and included in this process was the

20 establishment of a contingent valuation panel of

21 economic experts to evaluate the use of CVM in

22 determining none use values and provide comments to

23 NOAA. The panel members are Kenneth Arrow, do you

24 know no him?








88


1 A. Not personally. By representation.

2 Q. He's a recognized expert?

3 A. Well known expert.

4 Q. Robert Solow, S-O-L-O-W, do you know him?

5 A. Yes.

6 Q. He's a recognized expert?

7 A. Yes.

8 Q. Edward Leamer, do you know him?

9 A. Just by representation.

10 Q. You know of him?

11 A. Yes.

12 Q. He's a well recognized expert, right?

13 A. Yes.

14 MR. SWEDLOW: I guess I'll object. He is

15 asking for legal conclusions from this witness about

16 who they are.

17 MR. WAGNER: In your field as you

18 understand someone who has expertise.

19 THE COURT: Overruled. He means in

20 layman's terms.

21 Q. I'm sorry?

22 A. Yes.

23 Q. There you go. Paul Portney?

24 A. I don't that name.








89


1 Q. Do you know of him?

2 A. Of him.

3 Q. He has a representation of being an expert

4 as you understand that?

5 A. I have thin knowledge on his reputation.

6 Q. Roy Randner, R-A-N-D-N-E-R?

7 A. Again just a thin knowledge of his

8 reputaiton.

9 Q. Howard Schueman?

10 A. Schueman is very recognized.

11 Q. Very recognized as is survey expert?

12 A. Correct.

13 Q. He published with Stanley Presser?

14 A. I believe so.

15 Q. And of the panel throughout this effort

16 NOAA has attempted to provide an atmosphere which an

17 unbias academic analysis, a, the CVM could be

18 conducted. The panel received hundreds of pages of

19 comments concerning CVM and conducted a public

20 meeting to hear all sides of the issue. The panel

21 has finished its work and submitted its reported to

22 NOAA and it's included in this document as appendix

23 one and that's what we've included beginning on page

24 three. You see all that?








90


1 A. Yes.

2 Q. Would you turn to page fifteen, Exhibit

3 4348. Now, at the very top we see the survey

4 guidelines. And the NOAA panel says in this section

5 we try to lay down a fairly complete set of

6 guidelines compliance -- a complete set of

7 guidelines, compliance with which would define an

8 ideal contingent valuation survey. A CV survey does

9 not have to match each of these guidelines fully in

10 order to qualify as so source of reliable

11 information for a damage assessment process. Many

12 departures from the guidelines or even a single

13 serious deviation would, however, suggest

14 unreliability prima facie. Do you see that?

15 A. Yes.

16 Q. And you try to follow those standards,

17 then when you conduct contingent valuation surveys,

18 don't you?

19 A. No.

20 Q. I see. So on page fifteen the authors

21 then --

22 A. Excuse me.

23 Q. You said no, that's an answer. Thank you.

24 THE COURT: It's up to him.








91


1 Q. On page fifteen the authors give some

2 general guidelines that should be followed, you see

3 that, general guidelines?

4 A. Yes.

5 Q. Let's see how we do.

6 MR. SWEDLOW: If I can go back over

7 there?

8 THE COURT: Sure.

9 Q. Now one of the guidelines is to minimize

10 non-response, you see that?

11 A. Yes.

12 Q. High non-response rates would make the

13 survey results unliable, right?

14 A. (No response).

15 Q. Right?

16 A. Correct.

17 Q. This is because of something called

18 non-response is bias, correct?

19 A. Yes.

20 Q. Non-response bias refers to the

21 differences in attitudes and beliefs between those

22 who respond and those who don't respond to your

23 survey questions, correct?

24 A. Correct.








92


1 Q. Your non-response -- let me just make sure

2 we have this on here. Minimize non-response. Now

3 your non-response rate is 85 percent, wasn't it?

4 A. If you take into account all factors in

5 setting up the company core asset, the research

6 panel, the non-response is 85 percent. It was 35

7 percent for the last stage of the response.

8 Q. You are a member AAPOR?

9 A. Correct.

10 Q. And AAPOR defines guidelines as to how to

11 define response rates, right?

12 A. Yes, it does.

13 Q. If you follow the AAPOR the non-response

14 rate would be 85 percent?

15 A. We actually as a company have not

16 determined whether that AAPOR standard definition

17 applies entirely to our panel base design. There is

18 not a corollary to the standard definitions made in

19 AAPOR that applies to what we do.

20 Q. I see. You would agree, would you not,

21 that whenever you are doing surveys with your panel

22 that findings based upon probability samples with

23 high levels of overall non-response should be

24 confirmed by findings based on samples selected by








93


1 methods that produce more complete responses, don't

2 you, you would agree with that sentence?

3 A. Could you read that again?

4 Q. Sure. That findings based on probability

5 samples with high levels of overall non-response

6 should be confirmed by findings based on samples

7 selected by methods that produce more complete

8 response, you agree with that, wouldn't you?

9 A. In an ideal world you would do

10 supplemental surveys. In the real world it

11 frequently occurs that you don't.

12 Q. That's actually a principle that you

13 published in a JAMA Article in 2002, isn't it?

14 A. I believe you are talking about the

15 Schlenger article?

16 Q. I am.

17 A. Yes.

18 Q. Now the NOAA panel also addresses the need

19 to conduct personal interviews. If you are looking

20 on page fifteen again, which is 4348. You see

21 another right under minimize non-response. Personal

22 interview. You see that?

23 A. Yes.

24 Q. The panel says they believe it is unlikely








94


1 that reliable estimates of values can be elicited

2 with mail surveys, face to face surveys are

3 preferable although telephone interviews have some

4 advantages in terms of cost and centralized

5 supervision. Just so we understand, your survey did

6 not have personal interviews, did it?

7 A. No, it did not.

8 Q. Yours was a self-administered, right?

9 A. Web-enabled, self-administered.

10 Q. Self-administered over the Web?

11 A. Right.

12 Q. Flip to page 16. Right at the top there is

13 a discussion of careful pretesting the CV

14 questionnaire. You see that?

15 A. Yes.

16 Q. In general that's a principle you agree

17 with, right?

18 A. I agree with the principle of pretesting.

19 Q. The panel says, respondents at a CV survey

20 are ordinarily presented with a good deal of new and

21 often technical information well beyond that is

22 typical in most surveys. This requires very careful

23 pilot work and pretesting, plus evidence from the

24 final survey that the respondents understood and








95


1 accepted the main descriptions and questions

2 reasonably well, you see that?

3 A. Yes, I do.

4 Q. AAPOR agrees in their best practices they

5 believe to pretest?

6 A. Yes.

7 Q. And there are different types of pretest,

8 aren't there, one could conduct focus groups before

9 a survey?

10 A. Correct.

11 Q. You didn't do any focus groups here, did

12 you?

13 A. No, we did not.

14 Q. Another type of pretest might be to do

15 cognitive interviewing?

16 A. That's correct.

17 Q. You didn't do any of that here, did you?

18 A. No.

19 Q. Another type of analysis that one can do

20 about understanding would be behavior coding?

21 A. Correct.

22 Q. You didn't do that here, did you?

23 A. No.

24 Q. You didn't do any of that. Now, in fact,








96


1 when questions changed between the pretest and the

2 main test there was no further pretesting, was

3 there?

4 A. No, there was not.

5 Q. All right. And if we talk about the

6 damages questions of your survey, those begin at

7 question 12 in the main survey, right?

8 A. Correct.

9 Q. And there is an analog from question 11 of

10 the pretest, right?

11 A. Yes.

12 Q. The question 12 was very different than

13 the question 11, is that correct?

14 A. Yes.

15 Q. It involved some substantial changes,

16 didn't it?

17 A. My recollection is yes.

18 Q. For example question 11 in the pretest

19 refers to what somebody would have done in the past.

20 Where at question 12 in the main survey asks what a

21 person would do in a hypothetical situation in the

22 present?

23 A. Yes.

24 Q. That question was not pretested after it








97


1 had been redone, correct?

2 A. Correct.

3 Q. And in the pretest there weren't any of

4 the point estimate questions, were there?

5 A. No.

6 Q. Okay. So in the pretest either for the

7 just as harmful case or the could be more harmful

8 case, there weren't any questions about the point

9 estimates between 10 to 90 or 100 percent of whether

10 or not somebody would continue to buy a Marlboro

11 Lights at any price, were there?

12 A. No.

13 Q. So those questions weren't pretested at

14 all, were they?

15 A. No, they were not.

16 Q. On page 16 -- I'll just add in here no

17 willingness to pay a price point question is

18 pretested. Let's go back to page 16 there are some

19 other items that the panel recommends and I want to

20 make sure we talk about them real quickly. One is

21 called temporal averaging. You see that, it's second

22 from the bottom. The panel says, time dependent

23 measurement should be reduced by averaging across

24 independently drawn samples taken at different








98


1 points in time. A clear and substantial time trend

2 in the responses would cast doubt on the reliability

3 of the findings. You see that?

4 A. Yes.

5 Q. You didn't do that in this study, did you?

6 A. No.

7 Q. Let's take another one. No answer option.

8 Your no answer option should be explicitly allowed

9 in addition to yes or no vote options on the main

10 valuation question. Is there an explicit no answer

11 option?

12 A. No, there is not.

13 Q. Let's go to page 17 and pick another one.

14 Cross tabulations. Surveys should include a variety

15 of other questions to help to interpret the

16 responses to the primary valuation questions. The

17 final report should include summaries of willingness

18 to pay broken down by these categories. You didn't

19 cross tabulate your answers across the survey, did

20 you?

21 A. No, I did not.

22 Q. You didn't for example perform any

23 consistency check between the first part of the

24 interview and the second parted of the interview,








99


1 did you?

2 A. As part of the questionnaire?

3 Q. Let me ask you the question a little

4 differently. Maybe that wasn't clear, I apologize

5 for that. In the first part of the belief questions,

6 you were asking people about whether or not they had

7 a belief that descriptors had certain meanings and

8 whether that contributed to their purchase decision,

9 correct?

10 A. Correct.

11 Q. Okay. People had a chance to answer yes or

12 no to those, right?

13 A. Yes.

14 Q. And if they answered that they said yes to

15 those questions that would indicate that they had a

16 belief that the descriptor had the meaning and it

17 contributed to the purchase decision, correct?

18 A. That's correct.

19 Q. That would indicate the more yeses means

20 stronger the belief in what you were seeking to

21 understand, right?

22 A. I don't agree with that.

23 Q. Okay. Would you agree with me that if --

24 well, you don't agree with me that the answering the








100


1 yeses indicates a stronger belief in the meaning of

2 the descriptors?

3 MR. SWEDLOW: I object, we just changed the

4 question.

5 MR. WAGNER: It's cross examination.

6 THE COURT: Overruled.

7 A. In my opinion it's not an additive

8 function. You simply can't add up the number of

9 yeses between question one and question two and at

10 the end of the yes/no questions and come up with an

11 intensity scale of belief in the health --

12 Q. Fair enough, Dr. Dennis. For any question

13 on which a respondent answered yes did you correlate

14 that with their willingness to pay answers in the

15 second half of the survey?

16 A. No, I did not.

17 Q. The panel also says that there should be

18 checks on understanding and acceptance of the

19 questions, right?

20 A. Yes. I saw that in there.

21 Q. You are familiar with the con -- are you

22 familiar with the concept of something called a

23 dominance questions?

24 A. Yes.








101


1 Q. A dominance question, would you agree with

2 me, is one of the tests you can use to make sure

3 people are understanding the questions and

4 responding fairly, right?

5 A. Correct.

6 Q. And often times those are included in

7 surveys in order to check and make sure that people

8 are still awake, that they are reading the questions

9 and they are responding in the way the survey

10 designer intended for them to respond, correct?

11 A. Correct.

12 Q. Now, I want you to take a look in your

13 survey, you still have a copy of the survey?

14 A. Yes, I do.

15 Q. Take a look at question fifteen which is

16 on page 37 of the survey. Here you ask survey

17 respondents to "Suppose that there was another kind

18 of Marlboro Lights that could be more harmful than

19 Marlboro Reds because it has both an increased

20 potential for causing genetic and chromosomal damage

21 and contained higher levels of some of the most

22 toxic substances found in cigarette smoke as

23 compared to a Marlboro Reds cigarette. Which of the

24 following cigarettes would you choose? Remember that








102


1 both Marlboro Lights cigarettes taste exactly the

2 same and cost exactly the same." One: "The Marlboro

3 Lights cigarette that deliver less tar and is less

4 harmful or safer;" and two: "The Marlboro Lights

5 cigarette that could be more harmful than Marlboro

6 Reds because it had both an increased potential for

7 causing genetic and chromosomal damage and contained

8 higher levels of some of the most toxic substances

9 found in cigarette smoke as compared to Marlboro

10 Reds cigarette." Dr. Dennis, of the valid percent

11 tell me what percent of the sample chose the could

12 be more harmful cigarette?

13 A. Of those asked that question?

14 Q. Yeah.

15 A. 52 percent.

16 Q. 52 percent. Is question fifteen a

17 dominance question in your view?

18 A. Not in my view.

19 Q. Okay. You have any explanation why 52

20 percent of the people asked that question would

21 choose the more harmful cigarette if they had a

22 choice?

23 A. I would answer the question as follows:

24 You have to look at the subgroup that was actually








103


1 administered that question. That's the group that

2 initially told us they would prefer a more harmful

3 cigarette or just as harmful cigarette, question 12,

4 as you know. Secondly, it's a very small sample

5 size, talking about a total of 21 respondents, I

6 believe actually twenty respondents administered the

7 question. Therefore, the possibilities or the margin

8 of error there is very large. Thirdly, I would

9 suggest that you do have some respondents that may

10 not buy into the overall situation, if you will, of

11 this question. And that they are being in a sense

12 bad respondents and disagreeable, which is not an

13 uncommon situation in surveys to find a very small

14 fraction of your respondents give you responses that

15 are surprising. Public opinion surveys that are

16 frequently get 70 or 80 percent on a particular item

17 frequently invoke surprise because it's hard to get

18 100 percent agreement on anything, much less even 70

19 or 80 percent. So this particular result in context

20 of the respondents actually that were administered

21 this question is not surprising.

22 Q. One thing you mentioned, Dr. Dennis, is a

23 small sample size. You think it's inappropriate to

24 generalize for a sample size of twenty people,








104


1 right, from a sample size of twenty people, right?

2 A. Generally a sample size of twenty people

3 is hard to generalize.

4 Q. It would be worse if the sample size was

5 nine or five, wouldn't it?

6 A. The lower the number, yes, the less

7 confidence you have.

8 Q. Thank you very much. Let's talk about the

9 first part of the survey. Survey results can be

10 subject to bias, would you agree with me on that?

11 A. Yes.

12 Q. And in the field of survey research the

13 presence of bias detracts from validity, right?

14 A. Yes.

15 Q. Okay. Now, a bias survey is one that

16 through its questions leads people to give false or

17 inaccurate answers to the survey questions, right?

18 A. That's only one form.

19 Q. That's one form of bias. And bias, if you

20 were to give a definition to the Court, would be the

21 difference or the deltite (phonetic), I think is the

22 word you used, between the true measure or true

23 response and the response that's measured by the

24 survey, right?








105


1 A. That's correct.

2 Q. If one wanted to quantify the bias that's

3 how one would go -- that's the difference one would

4 be trying to look for?

5 A. Ideally, yes.

6 Q. There are different forms of bias that can

7 effect the survey's results?

8 A. That's right.

9 Q. Let's see if we can explain to the Court

10 what are the different types of bias. Okay. One

11 source of bias, you would agree with me, a

12 measurement error, correct?

13 A. Correct.

14 Q. Okay. That would be created by respondents

15 not understanding the question and/or not

16 understanding the response options so they end up

17 giving the wrong answer, right?

18 A. Yes.

19 Q. Another source of bias is something called

20 non-response bias, right?

21 A. Correct.

22 Q. We already talked a little bit about that,

23 didn't we? Non-response bias can occur as a result

24 of people refusing to participate in a survey,








106


1 right?

2 A. It's that and a compound of course if

3 there's attitudinal differences or other kinds of

4 differences between the non-responders and the

5 responders.

6 Q. So if people don't qualify or join the

7 panel and there are attitudinal differences that's

8 would lead to non-response bias?

9 A. Both those things must occur for

10 non-responsive bias.

11 Q. In other words, this bias exists if

12 there's a group of people not in the survey if they

13 have different attitudes and beliefs than those in

14 the survey then we need to know what those

15 differences are before we project to our population,

16 right?

17 A. Most the time it's not a knowable

18 proposition to know those differences.

19 Q. Okay. We also have a third type of bias

20 called none coverage bias?

21 A. Correct.

22 Q. This involves excluding from your sample

23 people that are eligible, should have received the

24 survey but didn't, correct?








107


1 A. Correct.

2 Q. And each of these forms of bias can

3 prevent the survey from presenting accurate

4 responses, right?

5 A. The summary results can be bias as a

6 result of any of these three.

7 Q. Okay. And as a result each of these forms

8 of bias can render a survey invalid, correct?

9 A. That's correct.

10 Q. All right. Let's each one in depth.

11 Measurement error. Error refers to bias that can

12 result from, for example, the wording of a question

13 in a survey?

14 A. That's correct.

15 Q. And when designing a survey you have to be

16 careful in drafting questions because the questions

17 can be worded in such a way that creates bias,

18 right?

19 A. Yes.

20 Q. Okay. The manner in which a question is

21 asked can greatly effect the survey results, right?

22 A. Sure.

23 Q. Would you agree with me that even

24 sometimes slight changes to the wording of a








108


1 question can lead to substantial changes in the

2 answers that survey respondents give?

3 A. That can happen.

4 Q. Okay. Let's talk about measurement

5 errors. Where does it come from? In this case you

6 were tasked with the responsibility of designing the

7 questions, you actually wrote the survey, right?

8 A. Correct.

9 Q. But that wasn't a job you performed on

10 your own, was it?

11 A. I don't understand.

12 Q. Isn't it true, sir, that you created the

13 survey in collaboration with plaintiffs' counsel?

14 A. No. The relationship was that plaintiffs'

15 counsel provided requirements, in some cases

16 scientific phrases or terms, as well as the overall

17 target population requirements for the project.

18 Q. You deny it was a collaboration?

19 A. I answered your question and described the

20 relationship.

21 Q. Sure. Do you deny it was a collaborative

22 effort between you and plaintiffs' counsel is my

23 question.



12 Q. You recall giving a deposition in this

13 case, right?

14 A. Yes, I do.

15 Q. You did gave that deposition on December

16 2nd, correct?

17 A. Correct.

18 Q. And if you turn to page 47 of the

19 deposition?

20 A. Yes.

21 Q. At line 19 question was, "Let me stop

22 there. Is it fair to describe the drafting of the

23 questionnaire as a collaboration between you and Mr.

24 Swedlow?" Answer, "I would characterize it as








110


1 myself as prime author and Mr. Swedlow as the

2 organizer-input from the other experts in terms of

3 providing me with requirements for Mr. Harris, you

4 mentioned. Then secondly, after the pretest

5 suggestions that came from Professor Presser." Skip

6 the next two paragraphs, although you can read this

7 if you like, I want to go to line 16 where you said,

8 "A collaboration? You used the word collaboration.

9 I think it was a collaboration in a sense that I

10 make suggestions about how to do this, I'm writing a

11 questionnaire, and I'm not unilaterally imposing it

12 on my customers. We have conversations about this

13 question wording, the series of questions, is going

14 to meet the objective or not. And then secondly, of

15 course, he has the other experts to consult with

16 too. So in that sense it's a collaboration, and

17 that's what this questionnaire is largely about.

18 It's a constellation of minds --"

19 A. Excuse me, you misread. "What

20 questionnaire design." Not this questionnaire, but

21 what.

22 Q. Thank you. "That's what questionnaire

23 design is largely about. It's a constellation of

24 minds brought together, each people having their own








111


1 specific strengths that they bring to the table.

2 And usually there's a core group or a single

3 individual that's actually bringing together the

4 questionnaire into one coherent piece. In that case

5 it was me."

6 MR. TILLERY: Your Honor, that's not

7 impeaching.

8 THE COURT: Is this intended for

9 impeachment?

10 Q. Is that sense in which you understood the

11 collaborative effort? Let me ask this question.

12 THE COURT: Wait a minute. Let him ask the

13 question and I'll --

14 Q. I want to know whether or not he is going

15 to agree with the characterization it was

16 collaboration with plaintiffs' counsel?


8 Q. Okay. Dr. Dennis, let me ask you.

9 THE COURT: Call it recollection of some

10 kind.

11 Q. All right. Does this refresh your

12 recollection.

13 THE COURT: Just go ahead.

14 Q. I'll ask the next question.

15 THE COURT: I don't want to waste your

16 time.

17 Q. Litigation is not the context in which you

18 ordinarily draft surveys, correct?

19 A. That's correct.

20 Q. So in your work you don't normally have

21 lawyers involved in the drafting of the

22 questionnaire, do you?

23 A. Lawyers involved in any sense, that's

24 correct.








113


1 Q. You usually work in collaboration with

2 other people who are experts in drafting survey

3 questionnaires, correct?

4 A. In many cases, yes.

5 Q. For example Dr. Fiscusi, he is an expert

6 you understand that he designs a lot of surveys?

7 A. He does.

8 Q. And you normally don't have lawyers who

9 have a financial interest in the outcome of the

10 survey helping you draft the questions, right?

11 A. I already answered the question. I

12 normally don't have lawyers involved in any form of

13 the questionnaires.

14 Q. Okay. Now, within a couple of days after

15 you were retained to work on this case you did write

16 the initial draft of the survey questions, right?

17 A. Yes.

18 Q. And then you passed the questions that

19 were drafted along to Mr. Swedlow, plaintiffs'

20 counsel, right?

21 A. Yes, I must have.

22 Q. And over the next three or four days you

23 and plaintiffs' counsel made changes to the

24 questionnaire, correct?








114


1 A. That's correct.

2 Q. Now, did you ask him by the way, if he had

3 any experience in designing surveys?

4 A. Did I ask Mr. Swedlow?

5 Q. Mr. Swedlow.

6 A. I would have no reason to ask him that. I

7 don't think I ever asked him that.

8 Q. Do you know whether he has any expertise

9 in drafting contingent valuation surveys?

10 A. I have no knowledge about Mr. Swedlow's

11 experience in drafting contingent valuation surveys.

12 Q. In these first few days you and Mr.

13 Swedlow, plaintiffs' counsel, sent at least a few

14 e-mails back and forth, correct?

15 A. I'm sure we sent some e-mails back and

16 forth.

17 Q. Most the time you talked with plaintiffs'

18 counsel over the telephone about ways to change the

19 questionnaire, correct?

20 A. We had phone calls where I discussed and

21 reviewed the status of the questionnaire.

22 Q. In fact in the first three or four days

23 you worked on the survey you had somewhere between

24 40 and 70 telephone conversations with Mr. Swedlow,








115


1 didn't you?

2 A. Yes I previous reported that.

3 Q. And over the three or four days of

4 drafting the initial version of the questionnaire

5 you made changes throughout the questionnaire to get

6 it ready to be pretested, right?

7 A. Yes.

8 Q. Your understanding from plaintiffs'

9 counsel is that Mr. Swedlow was sending the drafts

10 of the questions and discussing the questions to

11 other expert witnesses who were providing feedback

12 to him and he would then pass that feedback along to

13 you, right?

14 A. That's correct.

15 Q. Okay. And this took place over a two to

16 four day period, a very short time period, is that

17 right?

18 A. That's right.

19 Q. By the way, when you are conduct surveys

20 for or you are designing a survey for say the EPA,

21 you worked with Dr. Fiscusi, that's not done in two

22 to four days, is it?

23 A. No, that's a different type of project.

24 Q. Okay, I see. In fact, in this instance the








116


1 whole process was very compressed, you had to put

2 everything together in a seven to ten day period and

3 deliver the data?

4 A. That's right.

5 Q. Now, sometimes you suggested ways to

6 change the questionnaire and Mr. Swedlow rejected

7 your suggestions, wouldn't he?

8 A. I believe, yeah, that must have happened

9 where the scientific experts or others would make

10 recommendations, suggest what the requirements of

11 the survey had to be to meet the damages model or

12 whatever the evidence must be.

13 Q. Right. He told you that the scientists

14 wouldn't allow the changes to the question that you

15 suggested?

16 A. I think a conspicuous example is one about

17 the chromosomal damages and toxic substances, yes.

18 Q. On the other hand, when plaintiffs'

19 counsel, Mr. Swedlow, suggested changes to the

20 questionnaire, you would ordinarily implement those

21 suggestions?

22 A. Technical things that, you know, to meet

23 the requirements of the project from the scientists

24 then I would implement those changes.








117


1 Q. Okay. And the give and take that was going

2 on between you and Mr. Swedlow happened throughout

3 the time you were getting the questionnaire ready

4 for pretesting, right?

5 A. That's right.

6 Q. You kept making some incremental changes

7 to the survey as you got it ready to go between

8 those 40 to 70 telephone calls?

9 A. Correct.

10 Q. Just so we are clear on this, the only

11 person that you dealt with in creating the pretest

12 was Mr. Swedlow, plaintiff's counsel, right?

13 A. Right.

14 Q. He is the one that told you if your

15 suggestions for changing the survey before the

16 pretest would not be implemented, correct?

17 A. He was the only person I had conversations

18 with, so by definition.

19 Q. The two experts that he identified was

20 Stanley Presser and Jeffrey Harris, two of the

21 plaintiffs' experts in this case, right?

22 A. Those were the two experts that were

23 mentioned by name.

24 Q. You never spoke directly to either Dr.








118


1 Presser or Dr. Harris, correct?

2 A. That's correct.

3 Q. You just had to trust Mr. Swedlow that Dr.

4 Presser and Harris were suggesting changes that were

5 then being put in the pretest, correct?

6 A. I didn't make an assessment of trust or

7 distrust. I'm working with a customer.

8 Q. Are you aware, Dr. Dennis, that both Dr.

9 Presser and Dr. Harris have testified that they did

10 not comment on the wording of the questionnaire

11 prior to the time the wording was pretested?



11 Q. My question is whether or not Dr. Dennis

12 is aware that both Dr. Presser and Dr. Harris have

13 now testified that they in fact did not comment on

14 the wording of the questions prior to the time the

15 survey was pretested?

16 THE COURT: He can answer that.

17 A. Was I aware of this testimony?

18 Q. Right.

19 A. I'm not aware of that testimony.

20 Q. Let's see what they said.



12 Q. This is from Dr. Presser's deposition.

13 Page 65.



17 Q. Okay. Let me ask you this, did the

18 plaintiffs' counsel --



21 MR. WAGNER: That's fine, your Honor.

22 Q. Dr. Dennis, let me ask you this. Mr.

23 Swedlow did not identify to you any other experts

24 who were involved in suggesting changes to the








122


1 questionnaire, did he, other than Dr. Presser and

2 Dr. Harris?

3 A. Those are the only two names I'm aware of.

4 Q. Okay. And if Dr. Presser and Dr. Harris

5 did not comment on questions prior to the pretest

6 then so far as you are able to deduce plaintiffs'

7 counsel would have come up with all of the changes

8 without consulting with any of the experts, right?

9 A. Well, you know, my deposition -- will I

10 have an opportunity to --

11 Q. Mr. Swedlow will be able to come back.

12 Just answer my question?

13 A. Could you repeat your question?

14 Q. Sure. On the assumption that neither Dr.

15 Harris or Dr. Presser had any input in the wording

16 of the pretest then so far as you know all of the

17 changes that were suggested came solely and

18 exclusively from plaintiffs' counsel, correct?

19 A. Changes and objectives, correct.

20 Q. And apart from the screening questions

21 that we talked about, about who would get into the

22 survey, as you sit here today, you are not able to

23 reconstruct which questions, phrases or sentences in

24 this survey were suggested by plaintiffs' counsel








123


1 and which ones were not, can you?

2 A. No. I think there's a mischaracterization

3 or poor assumption or inaccurate assumption in the

4 question. I was not given questions by counsel.

5 There is no questions given to me by counsel. I

6 drafted the questions. What were delivered to me as

7 I went over three or four times in the deposition

8 were the requirements and specifications, the key

9 concepts that had to be embedded in order to satisfy

10 the legitimate needs of the project.

11 Q. I appreciate that, Dr. Dennis. You

12 already agreed with me sometimes even slight wording

13 changes can significantly effect the results,

14 correct?

15 A. Yes, that can happen.

16 Q. As you sit here today you are not able to

17 reconstruct which changes to the questions, which

18 phrases or even portions of sentences were suggested

19 by plaintiffs' counsel and which ones were not, can

20 you?

21 A. Sometimes I can. Other times I wouldn't be

22 able to make that recall.

23 Q. Okay. We are talking about a collaboration

24 with plaintiffs' counsel. Dr. Dennis, when you were








124


1 working through the pretest and ultimately towards

2 the main survey, you didn't retain any of the

3 earlier drafts or any record that would indicate how

4 the questionnaire changed as a result of yours and

5 plaintiffs' counsel comments, correct?

6 A. That's correct.

7 Q. Okay. In fact, isn't it true, Dr. Dennis,

8 you did whatever you could to make sure there was no

9 written record of how you and the plaintiffs'

10 changed the questionnaire?

11 A. That's untrue.

12 Q. Let's go over it specifically. Over the

13 course of your work on the survey did you send some

14 e-mails to plaintiffs' counsel, right?

15 A. Correct.

16 Q. But those e-mails were very limited, they

17 related to contract matters, not anything

18 substantive which related to the survey, right?

19 A. That's my recollection, yes.

20 Q. And so although there were some e-mails

21 most of the time you communicated orally with

22 plaintiffs' counsel over the phone, which doesn't

23 leave a written record, right?

24 A. That's correct.








125


1 Q. And in fact, when it comes to the survey

2 rather than e-mailing or faxing copies of the

3 questionnaire back and forth with the changes you

4 and the plaintiffs' counsel would talk on the

5 telephone and then each of you would type the

6 changes into your own separate documents, right?

7 A. Well, I was the owner of the document I

8 would make changes to the document. We talk about

9 them on the phone and I was making my suggestions

10 and recommendations. You have the full deposition

11 record. And I made those changes directly to the

12 document.

13 Q. And on your computer when your are making

14 changes to the document every time you and Mr.

15 Swedlow made changes to it you saved the new version

16 right on top of the old version so the old version

17 no longer existed, right?

18 A. I think that's a punitive way to put it. I

19 simply saved the document.

20 Q. You saved it over the old version, so we

21 couldn't receive the old version?

22 A. You make it sound like it was an

23 intentional act. I just saved the document.

24 Q. I'm not asking about intent I just want to








126


1 know what happened.

2 THE COURT: Just answer the question.

3 A. I would save that document.

4 Q. So there would be no record, no historical

5 record of the changes then, right?

6 A. That's correct.

7 Q. Now let's get a sense of the timing here.

8 There became a point in time in the drafting process

9 when you were preparing to get the initial version

10 of the survey ready for the pretest, right?

11 A. Yes.

12 Q. And after the three or four days of

13 drafting and 40 to 70 phone calls you had addressed

14 all of the comments on the questionnaire that you

15 would have received from plaintiffs' counsel, right?

16 A. Correct.

17 Q. And after that you pretested the initial

18 version of the survey, right?

19 A. Yes.

20 Q. You two purposes for the pretest?

21 A. Yes.

22 Q. First, you wanted to make sure that enough

23 of the smokers in the data base where going to be

24 eligible to take this survey, right?








127


1 A. That's correct.

2 Q. Second, you wanted to make sure the skip

3 patterns that you programmed into the computer were

4 operating correctly, right?

5 A. That's correct.

6 Q. And third, you wanted to make sure that

7 there weren't any questions that were problematic

8 and needed to be redesigned?

9 A. That's correct.

10 Q. And one way you can determine whether

11 questions are problematic is by looking at whether

12 people skipped the question or broke off the survey

13 all together?

14 A. That's correct.

15 Q. Okay. And you ran the pretest and found

16 you did in fact have enough smokers in your data

17 base who would be eligible for the survey, right?

18 A. Yes. That wasn't the objective. I wanted

19 to confirm that the rate was what I thought it was

20 going to be.

21 Q. And you confirmed that, right?

22 A. Yes.

23 Q. The pretest also confirmed that with one

24 minor exception the skip patterns you programmed








128


1 into the computer were working correctly?

2 A. Yes.

3 Q. And finally the pretest showed that the

4 respondents were not skipping the questions or

5 breaking off from the survey?

6 A. Right.

7 Q. In your mind nothing in the pretest

8 indicated that any of the survey questions were

9 problematic or needed to be changed, right?

10 A. Based on the examination of the pretest

11 data, but based on other sources of examining the

12 questionnaire more closely there were ideas about

13 how to change the questionnaire.

14 Q. Oh, yeah, we are going to get to those.

15 Based upon the pretest data nothing was suggested by

16 the pretest data, nothing indicated that the survey

17 questions were problematic or needed to be changed,

18 right?

19 A. That's right.

20 Q. So after the pretest you then calculated

21 people's answers to the questions, sent the answers

22 to Mr. Swedlow, plaintiffs' counsel, right?

23 A. Correct.

24 Q. At that time -- at the time that you








129


1 conducted your pretest you already implemented all

2 of the plaintiffs' counsels comments on the survey

3 questionnaire, right?

4 A. I had finished the drafting of the

5 questionnaire and implemented that. I --

6 Q. Were you finished with your answer? If I

7 interrupt it's only because I am trying to go fast I

8 don't mean to do that.

9 A. I stutter.

10 Q. Let me ask you the question this way.

11 Everything that you and Mr. Swedlow wanted was in

12 the pretest, wasn't it?

13 A. All the questions that had been finalized

14 were part of the pretest questionnaire.

15 Q. I asked you a very specific question, I'll

16 ask it again. I want to make sure you got it. Isn't

17 it true that everything that you and Mr. Swedlow

18 wanted in the pretest got in that pretest?

19 A. I think the way that you are phrasing the

20 question overstates the extent of the collaboration.

21 Q. Take a look at our deposition again. Page

22 74.


21 Q. Okay. You received additional -- did you

22 receive -- strike that. Let me start that again.

23 Within a very narrow slice of time, Dr. Dennis,

24 during the time when you began the pretest and the








131


1 time when the pretest concluded, so you can do the

2 results, isn't it true you received no additional

3 comments on the questionnaire from plaintiffs'

4 counsel?

5 A. I have to go back to see what I said in my

6 deposition.

7 THE COURT: If you recall. If you recall.

8 If you don't recall say so.

9 A. I don't recall exactly.

10 Q. Take a look at page 75 and see if we can

11 refresh your recollection this time.

12 A. Can I hear the question again?

13 Q. I want to know whether in the narrow

14 window of time whether you received any suggestions

15 for changing the questionnaire while the pretest was

16 being fielded. I ask Dr. Dennis to look at page 75

17 of his transcript, lines 6 through 15. Begin the

18 question on line 7. "Between the time when you

19 started the pretesting process and when you produced

20 the data set, the weights, and distributions from

21 the pretesting --" you said "right." The question

22 continues "-- within that little window of time did

23 you receive any suggestions for changing the

24 questionnaire?" Answer, "No, I don't think so."








132


1 Does that refresh your recollection then as to

2 whether within that window of time you received any

3 suggestions for changing the questionnaire?

4 A. I trust my recollection at the time of the

5 deposition.

6 Q. So it was only after you conducted the

7 pretest and after you sent the survey results to

8 plaintiffs' counsel that plaintiffs' counsel then

9 suggested additional changes to the wording of the

10 questions, right?

11 A. Suggested, you know, I started to receive

12 input at that point.

13 Q. Okay. After the pretest you and the

14 plaintiffs' counsel had a discussion as to whether

15 or not the survey was meeting plaintiffs' counsel's

16 objectives, right?

17 A. Right.

18 Q. And during that conversation and

19 subsequent discussions plaintiffs' counsel did in

20 fact suggest further changes to the questionnaire,

21 right?

22 A. Yes.

23 Q. And as a result you made some systematic

24 changes to the questions regarding survey








133


1 respondents' beliefs and reasons for purchasing,

2 right?

3 A. Particularly part two of the questionnaire

4 are you referring?

5 Q. I'm focusing on part one, the beliefs and

6 attitudes part. The part about -- right, the part

7 about the beliefs and attitudes?

8 A. I don't remember right now the changes in

9 between the pretest and main. I have to look at all

10 my notes.

11 Q. Why don't you look at page 73 of your

12 deposition. And it begins at line -- begins at line

13 17 and the question carries over to 74 and see if

14 that refreshes your recollection as it whether or

15 not after the pretest you had discussions which lead

16 to systematically making changes to the attitudes

17 and beliefs questions?

18 A. Oh, right. Uh huh.

19 Q. So the answer is now yes, right?

20 A. I recall. I'm not sure if it's in the

21 deposition without reading further. I do recall

22 conversations about the first few questions about

23 that initial anchor.

24 Q. If you see on page 74 line two and three








134


1 "systematically making changes to the questions."

2 Does that refresh your recollection?

3 MR. SWEDLOW: I object, if the question is

4 does he see those words, he can answer that, but I

5 think he provided an answer that doesn't need

6 refreshing.

7 THE COURT: Why don't you do it that way.

8 Q. Do you see the words?

9 A. The word systematic?

10 MR. SWEDLOW: Line two.

11 THE COURT: You know what, keep going with

12 what you are doing. I don't want you to waste your

13 time.

14 Q. Does that refresh your recollection, sir,

15 as to whether or not you systematically made changes

16 on the question wording?

17 A. I know what it's referring to.

18 Q. I know what it's referring. My question

19 is, whether or not your recollection is refreshed

20 about whether or not you systematically made changes

21 to that set of questions, that's all I want to

22 know.

23 A. Yes.

24 Q. Systematic changes.








135


1 A. It's --

2 Q. Let's see what effect those had. Before we

3 get specific let me ask you this. You talked in your

4 directed testimony about valid percent, you recall

5 that?

6 A. In my testimony today?

7 Q. Yes.

8 A. Yes.

9 Q. Valid percent, there are also results that

10 your report in your survey for something called

11 percent?

12 A. That's correct.

13 Q. Percent actually refers to the percentage

14 of the total sample, right?

15 A. That's correct.

16 Q. So if the Court wanted to know what

17 percentage of the total sample was responding

18 affirmatively or negatively to the questions the

19 Court would look at the percent column?

20 A. No, they would look at the valid percent

21 column.

22 THE COURT: At the what?

23 A. The Court would look at the valid percent

24 column. That's the column that would show the








136


1 respondents that actually answered the question.

2 Q. Right. The valid percent refers to a

3 percentage of people that were actually asked a

4 particular question in the beliefs and attitudes

5 survey?

6 A. Yes.

7 Q. But you began the survey by asking people

8 whether or not they had any recollection at all as

9 to why they were purchasing Marlboro Lights when

10 they first made a purchase, right?

11 A. Yes.

12 Q. And if they answered no to that, you just

13 excluded them from the following questions, right?

14 A. Right.

15 Q. So my question to you is, if the Judge

16 wanted to know the percent of the total sample that

17 responded affirmatively to one of our other

18 questions the Court would look to the percent?

19 A. You can't respond to an affirmative

20 questions that is not asked.

21 Q. Let me ask you the question this way. If

22 you asked the question then -- you asked the

23 questions in the attitudes and beliefs portion by

24 giving a yes or no option?








137


1 A. Correct.

2 Q. If you asked them about don't know because

3 I don't recall why I bought this then there would be

4 a don't know option?

5 A. That's right.

6 Q. If we added them up all the answers for

7 yes, no, I don't know, the percentage of yes would

8 be the total sample, right?

9 A. In that case if all persons were

10 administered that question then the percent column

11 would show the percent that answered affirmatively.

12 But that's not what happened.

13 Q. I understand that. Let's see if we can see

14 what effect, I just want to ask you questions about

15 the percent sample. So if you will stay with me. I

16 want to know about the percentage from the total

17 sample who responded affirmatively to a question,

18 you with me?

19 A. Yes, I am.

20 Q. Let's take a look and compare the

21 questions from the pretest and then the questions

22 from the main survey after the systematic changes

23 and see if that effected the results. Why don't you

24 turn to page 12 of your expert report and look at








138


1 question 2. That's the first attitude and belief

2 question, right?

3 A. Yes.

4 Q. In question 2 you asked, "When you first

5 chose to smoke Marlboro Lights, did you think that

6 the words 'lowered tar and nicotine' that appeared

7 on the cigarette packages of Marlboro Lights meant

8 that a Marlboro Lights cigarette was lower in tar

9 and nicotine when you smoked it than a Marlboro Reds

10 cigarette?"

11 A. That's correct.

12 Q. And what percent of the total sample,

13 percent column answered yes?

14 A. It was at 50 percent for the percent

15 column.

16 Q. 50.3 percent. If you look at the main

17 survey that's at page 26. You ask the analog, right?

18 A. Correct.

19 Q. And that's after the systematic changes.

20 What happens to the total percent?

21 A. It's 63.9 percent.

22 Q. That's 13.6 percent increase, right?

23 A. Yes, you are applying a base that doesn't

24 make sense.








139


1 Q. That's what we just went over. We are

2 focusing of the percent of total sample. Percent

3 column this is what happened between the pretest and

4 main survey, right? Is that what happened if you

5 look at that column?

6 A. The labeling is not complete on your

7 table.

8 Q. All I'm asking you, sir, is whether or not

9 that's what happened if you look at the percent of

10 the total sample as you reported as percent in your

11 expert report, is that what happened?

12 A. I just don't think that's a valid way to

13 present the results.

14 Q. All I am asking you is whether or not that

15 happened?

16 A. Those are the numbers from the percent

17 column.

18 Q. Let's look at question three of the

19 pretest. Question three you ask, "When you first

20 chose to smoke Marlboro Lights, did you think that

21 the words 'lowered tar and nicotine' that appeared

22 on the cigarette package of Marlboro Lights meant

23 that a Marlboro Lights cigarette was less harmful in

24 some way than a Marlboro Reds cigarette?" Right?








140


1 A. Right.

2 Q. And then the percent of the total sample

3 affirmatively was 41.9 percent?

4 A. Right.

5 Q. And if we go to the main survey after we

6 make the systematic changes which is on page 27 the

7 analog is question four, correct?

8 A. Yes.

9 Q. And when you ask the question after the

10 systematic changes then the valid percent goes --

11 the total percent goes up, the percent of the total

12 sample answering affirmatively is 52.1 percent,

13 correct?

14 A. Yes.

15 Q. That's an increase of 10.2 percent,

16 correct?

17 THE COURT: Excuse me, he is testifying to

18 the pretest.

19 MR. WAGNER: That's correct.

20 THE COURT:. You are introducing that at

21 this time to compare the pretest with this survey?

22 MR. WAGNER: To show what happens.

23 Q. If we look at question four of the

24 pretest. You have asked whether or not the lower tar








141


1 and nicotine contributed to a purchase decision,

2 right?

3 A. To the initial decision to smoke, not

4 purchase decision.

5 Q. Oh, I'm sorry. To the initial decision to

6 smoke. That's the question you asked?

7 A. Correct.

8 Q. And the percent of total sample answered

9 affirmatively to that question was 39.3 percent,

10 right?

11 A. Correct.

12 Q. And if we go then to question five which

13 is the analog in the main survey which is at page

14 28, that is the analog?

15 A. Question five.

16 Q. And when you asked that question after the

17 systematic changes the percent of the total sample

18 responding affirmatively goes up to 50.1 percent?

19 A. Yes.

20 Q. That's an increase of 10.8 percent,

21 correct?

22 A. Yes. I'm looking at all the numbers. Yes,

23 those are the right numbers.

24 Q. Let's take a look at the next question,








142


1 question six. You switched over to Lights and you

2 asked whether or not the word Lights in the name

3 Marlboro Lights meant that a light cigarette was

4 lower in tar and nicotine when you smoked it than a

5 Marlboro Reds?

6 A. Right.

7 Q. When asked that the percentage of the

8 total responding affirmatively was 49.9 percent,

9 right?

10 A. Correct.

11 Q. An analog to that question in the main

12 survey is question number seven, correct?

13 A. Yes.

14 Q. And after the systematic changes to that

15 question the percents of total sample responding

16 affirmatively to that went up to 59.4 percent,

17 right?

18 A. No.

19 Q. Percent of the total sample?

20 A. I am objecting to the first part of your

21 sentence. You said after the systematic changes to

22 that question. Were there any systematic changes to

23 that question? I think systematic implies changes

24 across a broad number of questions. It cannot apply








143


1 to a single question.

2 Q. After you asked question seven of the main

3 survey did the percent of the total responding

4 affirmatively go to 59.4 percent?

5 A. Yes.

6 Q. Okay. That's an increase of 9.5 percent,

7 right?

8 A. Right.

9 Q. Let's look at question seven from the

10 pretest, you asked whether or not the association of

11 Lights with lower tar and nicotine contributed to

12 their initial decision to smoke Marlboro Lights,

13 right?

14 A. Right.

15 Q. And when that question was asked the

16 percents of the total sample responding

17 affirmatively was 47.5 percent, right?

18 A. That's correct.

19 Q. Question eight is the analog when you go

20 to the main survey, right?

21 A. Right.

22 Q. When you reask the question written in a

23 different way the percent of the total sample

24 answering affirmatively now went up to 55.2 percent,








144


1 correct?

2 A. It's not a causal connection between the

3 way the question was worded versus the result.

4 Q. Chronologically that's what happened,

5 right? You changed the question asked in the main

6 survey and got a different answer?

7 A. My opinion is changing the question did

8 not result in that increase.

9 Q. I understand that. Sir, my question is

10 whether or not after you asked the question in eight

11 in the main survey which is different from question

12 seven in the pretest where the percentage of total

13 sample responding affirmatively was 55.22 percent?

14 A. That's correct.

15 Q. Okay. That's an increase of 7.7 percent,

16 right?

17 A. Right.

18 Q. Question eight in the pretest asked

19 whether when people first chose to smoke Marlboro

20 Lights they thought the word Lights meant less

21 harmful in some way than a Marlboro Reds cigarette,

22 is that correct?

23 A. Yes.

24 Q. The percent of the total sample responding








145


1 affirmatively to that question was 45 percent,

2 right?

3 A. Right.

4 Q. And then the analog to this question is

5 question nine of the main survey?

6 A. Right.

7 Q. And after question nine was asked, which

8 is a different form than the pretest, the percent of

9 the total sample responding affirmatively went up to

10 54.4 percent, right?

11 A. Correct.

12 Q. That's an increase of 9.4 percent, right?

13 A. Right.

14 Q. All right. One more. Question nine whether

15 or not Lights contributed to the purchase, right?

16 A. Right.

17 Q. And the percent of the total sampling

18 answering affirmatively was 39.6 percent?

19 A. Right.

20 Q. And the analog for that is question ten in

21 the main survey, right?

22 A. Uh huh.

23 Q. So that question when asked a little

24 differently in the main survey the percent of the








146


1 total sample answering affirmatively went up to 51.8

2 percent, didn't it?

3 A. Yes.

4 Q. That's an increase of 12.2 percent, right?

5 A. Yes.

6 Q. Let me ask you one more thing with valid

7 percents. Question four in the main survey asks

8 customers whether or not -- asks survey respondents

9 whether lower tar and nicotine means less harmful,

10 right?

11 A. Yes.

12 Q. And even on the valid percent the answer

13 you got was 68 percent, right?

14 A. Yes.

15 Q. That's not a 100 percent?

16 A. No, it's 68 percent.

17 Q. Question nine asked survey respondents

18 whether Lights means less harmful, doesn't it?

19 A. Yes.

20 Q. And you didn't get 100 percent response to

21 that question either, did you, even if you wanted to

22 look at a valid percent which is asking a smaller

23 sample of people, right?

24 A. Right. 70.9.








147


1 Q. Not 100 percent, right?

2 A. Right.

3 THE COURT: You have five minutes.

4 Q. All the questions in the attitude and

5 beliefs part of the question were framed as yes/no

6 questions, correct?

7 A. With the exception of the open ended

8 questions, that's correct.

9 Q. You are familiar with the concept of

10 acquiescence bias?

11 A. Yes.

12 Q. Did you attempt to do anything in your

13 survey in order lessen the effect of acquiescence

14 bias and the answers given on the first part of the

15 survey?

16 A. Measurement, no, there was no measurement.

17 Q. And in questions -- in the questions in

18 the first part of the survey where you also asked

19 yes/no questions, the answer that was favorable to

20 the plaintiff was always the first response, right?

21 A. I don't recognize your concept of

22 favorable to -- favorable to whom?

23 Q. To the plaintiffs?

24 A. No. You are making suppositions about my








148


1 motivations.

2 Q. Not your motivations. Was the question --

3 in the order of the responses that was given?

4 A. I'm not a --

5 Q. The first response was always the one that

6 would say the descriptive had the particular meaning

7 suggested in the question?

8 A. I can't answer your question. I did not

9 design the questionnaire from the point of view of

10 knowing, having knowledge.

11 Q. When I reasked the question I didn't you

12 ask about plaintiff --


14 Q. Let me ask you this, Dr. Dennis --

15 THE COURT: You have an extra minute.

16 Q. In the attitudes and beliefs questions

17 when you were asking survey respondents what

18 descriptives meant you always suggested a meaning to

19 them in the questions, right?

20 A. Yes.

21 Q. And the meaning that was suggested you

22 always had to agree or answer yes as the first

23 response, right?

24 A. That's correct.








151


1 MR. WAGNER: How much time do you have,

2 your Honor?

3 THE COURT: I'll give you two and a half

4 more minutes.

5 Q. Dr. Dennis, in the attitude and beliefs

6 questions you did not offer survey respondents

7 "don't know" as an option, did you?

8 A. That's correct.

9 Q. Are you familiar with the work in the

10 field as to whether or not offering a "don't know"

11 response or failure to use a "don't know" response

12 can create a bias?

13 A. I'm aware of my own research and the

14 research of others in the field.

15 Q. Okay. And are you familiar with -- Dr.

16 Dennis, isn't is it true that some people when they

17 first started smoking Marlboro Lights made not have

18 seen the pack of cigarettes, they may have bummed a

19 cigarette off a friend?

20 MR. SWEDLOW: I object. This is outside the

21 scope of the proffered expertise of this witness.

22 MR. WAGNER: Give a tiny bit of leeway?

23 A. As a practitioner and non-smoker I can't

24 answer that question.








152


1 Q. Your survey questions give people a word

2 that appears, a word or phrase that appears on a

3 pack of Marlboro Lights and suggests a particular

4 meaning to it, right?

5 A. Yes.

6 Q. But other than lower tar and nicotine and

7 Lights you didn't ask people about any of the other

8 words and phrases that appear on those packs?

9 A. Yes.

10 Q. You are aware that ever pack of Marlboro

11 Lights cigarettes ever sold in Illinois has the

12 Surgeon General warnings, right?

13 A. I assume that's the case.

14 Q. You are aware also that the Surgeon

15 General warnings is the same as the one that appears

16 on the packs of Marlboro Reds, right?

17 A. Yes.

18 Q. But in your survey you never asked any

19 respondents what they thought the Surgeon General

20 warnings meant, did you?

21 A. No, because of the same between the two

22 products.

23 Q. Okay. Are you aware, Dr. Dennis,

24 statistical programs and packages that evaluate the








153


1 complexity of the wording of questions?

2 A. Such as reading levels?

3 Q. Yes.

4 A. Yes.

5 Q. Did you perform any of those tests on the

6 survey questions that you asked in the survey?

7 A. No.

8 Q. Okay. But you tried to, your panel of

9 Knowledge Networks respondents generally read from

10 eighth to tenth grade level?

11 A. No, I wouldn't say generally. The

12 Knowledge Networks panel is representative of the

13 United States for reading levels. I assume it's

14 higher than eighth to tenth grade level.

15 Q. When you are administering Internet based

16 surveys you tend to try to stay at about eighth to

17 tenth grade reading level?

18 A. I think that's standard practice cross all

19 kinds of surveys whether Internet or none Internet

20 surveys.

21 THE COURT: Why don't you ask one more

22 question.

23 Q. This survey, isn't it true, Dr. Dennis,

24 that the plaintiffs' asked you for a large sample of








154


1 Illinois residents but you were unable to do that?

2 A. They requested an Illinois base study to

3 use Illinois residents, but we could not do that.


5 REDIRECT EXAMINATION

6 QUESTIONS BY MR. SWEDLOW:

7 Q. With respect to the don't know response,

8 Dr. Dennis, have you done any research as to whether

9 or not the Knowledge Networks survey methodology or

10 any survey methodology is bias by the lack of a

11 don't know response?

12 A. Yes.

13 Q. What are your findings?

14 A. The major research findings is that it

15 doesn't make any difference whether you include a

16 don't know response or not. People who choose don't

17 know options are evenly distributed across the other

18 valid response options at the end of the day.

19 Q. ? While that is coming up, do you have an

20 opinion as to whether or not the Knowledge Networks

21 survey methodology and this survey in this case

22 suffers from any non-response bias based upon the

23 impaneling method?

24 A. In fact the method research that we have








156


1 done shows our particular methodology is very

2 comparable to telephone and other surveys that have

3 high response rates.

4 Q. Do you believe the wording of the

5 questions that you wrote makes the survey suffer

6 from measurement error?

7 A. No. I think the surveys are well-crafted

8 questions, well-crafted and not subject to any

9 significant measurement error to a respondent

10 difficulty in reading questions.

11 Q. The National Oceanic and Atmospheric

12 Administration or NOAA panel what where those

13 contingent valuation surveys meant to capture?

14 A. That particular panel in my opinion and my

15 understanding was directed to a specific species of

16 contingent valuation methodology studies where the

17 research topics were very complex, not involving

18 personal issues, often times involving very remote

19 hypothetical situations, and in my opinion not

20 applicable to personal risks decisions that people

21 take or consumer decisions such as this one in terms

22 of smoking cigarettes.

23 Q. You were asked whether or not these

24 questions had explicit "no answer" option. And you








157


1 said no there was no explicit "no answer" option.

2 Was there a way for the survey respondent not to

3 answer a particular question?

4 A. Yes, there was. There's an implied don't

5 know because respondents can skip or refuse to

6 answer any question. Simply press the continue

7 button to go to the next screen.

8 Q. This is one of the valuation questions

9 that we were referring to, you were asked several

10 times whether there was a dollar and cents valuation

11 provided. Is there an valuation for the Marlboro

12 Lights that delivers less tar and less harmful in

13 all of these questions?

14 A. A valuation?

15 Q. Is there a value placed on that for this

16 comparative valuation?

17 A. There is a question being asked for people

18 to compare two products and they are being asked

19 which one they would buy.

20 Q. For -- that's the wrong one. In the second

21 half of the survey is the price of the less harmful

22 Marlboro Lights anchored to the price that the

23 person usually pays in all of the questions?

24 A. That's correct.








158


1 Q. Does that provided for in your opinion as

2 an expert survey research practitioner, does that

3 provide an anchor or reference point for the

4 respondent to value the difference between these two

5 products?

6 A. In my opinion it's a better anchor than

7 making up a number about what they are going to pay

8 or multiplying current price. Anchoring in terms of

9 what people currently pay is something that

10 respondents understand.

11 Q. And why is it better to anchor it in what

12 they currently pay then to make up some price?

13 A. I think that it could create a more

14 cognitively challenging comparison for the

15 respondent.

16 Q. Do you believe that this survey, the first

17 half or the second half of the main study suffers

18 from acquiescence bias?

19 A. In my opinion it does not for the simple

20 reason that this is a self-administered form data

21 questions. There is no interviewer involved, whether

22 on telephone or in person. There's simply not

23 another party to please is what you see in typical

24 acquiescence bias examples.








159


1 Q. This question here we are discussing

2 acquiescence bias, I'm not asking you to determine

3 whether it's beneficial to the plaintiffs or

4 beneficial to the defendants, but this question asks

5 if there is any lower price that somebody would pay

6 for the less harmful product -- for the just as

7 harmful product. If the answer to this question is

8 yes, what will be the overall valuation of the harm

9 attributed in this case?

10 A. If they say yes to this question they are

11 placing some value on the just as harmful cigarette.

12 Q. Does that lead to a decrease in value for

13 the harm attribute?

14 A. It does.

15 Q. If the acquiescence --

16 THE COURT: Did he answer that?

17 A. Yes.

18 Q. If the acquiescence bias was applied to

19 this question, as you were discussing with

20 defendant's counsel, would it lead to more yeses or

21 more nos?

22 A. That's exactly right. You would expect

23 more yeses if acquiescence bias were the case or if

24 respondents were simply not paying attention, simply








160


1 selecting the yes time after time.

2 Q. Do you believe any of the questions in

3 this survey suffer from that acquiescence or not

4 paying attention or yes bias?

5 A. No, I do not.

6 Q. You were asked to compare a bunch of

7 percentages that were not valid percentages, but

8 percentages of people from the total respondent

9 pool, remember that?

10 A. Yes.

11 Q. Do you think that comparison has any

12 meaning for purposes analyzing the data?

13 A. No.

14 Q. If you were really going to compare the

15 percentage of people who were asked survey questions

16 to the total of people provided an opportunity to

17 take the survey, what number would we use from this

18 chart, meaning if we got respondents who said yes

19 compared to the total people who took any questions

20 in the survey, which would we use?

21 A. The 276 is your denominator.

22 Q. What if we wanted to include people who

23 were asked screening survey questions?

24 A. That's the 1779.








161


1 Q. Let's assume everybody who was asked the

2 question, 100 percent said yes to a question. But

3 you included everyone who answered screening survey,

4 whether or not they were asked the main interview.

5 What's the highest possible percent you could have

6 for any particular question?

7 A. Probably about 12 percent.

8 Q. Wouldn't it be 15.5?

9 A. Right.

10 Q. I didn't mean to do the math on the fly.

11 If you use those kinds of percentages, include

12 people who weren't asked a question, the highest you

13 could possibly get is 15.5 percent, isn't this

14 correct?

15 A. That's correct.

16 Q. When you compare the results of that first

17 part of the survey that you were being asked to

18 compare and you used the valid percentage, are there

19 any significant differences in the percentages

20 between the people who said yes in the first half

21 and the people who said yes in the first half in the

22 final survey?

23 A. If you look at the valid percentages

24 between the pretest and the main survey you see








162


1 relatively small differences. You see the larger

2 differences only when you use the percent column.

3 Q. I asked you this question prior to

4 responding to the Judge's point that you are also

5 going to be continuing as a professional after this

6 case. Under oath, Dr. Dennis, I'm asking you this

7 question, were any of the changes made from the

8 pretest to the main interview meant -- made to bias

9 or change the direction of any of the results of

10 this survey?

11 A. Absolutely not.

12 Q. Did you apply all of your knowledge and

13 powers based upon 13 years of survey research

14 expertise to draft as an objective and unbias survey

15 as you possibly could?

16 A. It's an unqualified yes.

17 Q. I would like to refresh your recollection

18 as to the way the other experts were involved in

19 this case by reading you an answer from your

20 deposition starting on page --



5 Q. Do you recall specifically whether the

6 other experts involved in this case participated

7 after the pretest or before the pretest?

8 A. My recollection is they participated after

9 the pretest was done.

10 Q. I don't need to refresh your recollection.

11 You were asked questions about whether or not this

12 process was a collaboration between me and you. And

13 you and I both know that process, but describing it

14 for the Court, would you say that the process was

15 unusual or uncharacteristic as compared to your

16 other clients, other than the fact that I'm an

17 attorney?

18 A. The unusual part would be the speed. The

19 questionnaire needed to be developed very very

20 quickly. Certainly I worked with media and other

21 clients even in commercial market research clients

22 where I developed a questionnaire very quickly with

23 mostly telephone base exchanges, for that part of

24 the work that I do it wasn't unusual.








164


1 Q. Going through without doing it

2 individually, based upon the differences between

3 what NOAA contingent valuation surveys are intended

4 to measure and what you were trying to measure with

5 your valuation survey, do you believe those criteria

6 would apply to your survey in this case?

7 A. No, I do not.

8 Q. Did anything that was refreshed in terms

9 of your recollection or mentioned on

10 cross-examination change any of your opinions in

11 this case as to the reliability and trustworthiness

12 of the survey you are offering into evidence today?

13 A. No. I stand by my deposition.

14 MR. SWEDLOW: I have no further questions.

15 THE COURT: Recross?

16 MR. WAGNER: No, your Honor. I'm fine.

17 THE COURT: We are done. Thank you, doctor.

18 MR. TILLERY: We move the admission of the

19 Exhibit 98.