Tobacco Documents Online

PROJECT I.~H;AT - PART l CLUSTER PROFILES OF U.K. Y~XI.E GXCF~'3',S AND TItEIR GENERAl, SMOKING 1L~BI'I'S REPORT NO. RD.1229-R 10.7.1975 ISSUED BY: C.I. Ayres DISTRIBUTION: Dr. S.J. Green Dr. I.W. llughes Dr. R.A. Sanfo,:d R.M. Cibb, Esq. F,.S. Wade, l'sq. R.C. Ilicho11,~;, I'sq. }iCtr i{. .~C, LL!;F: I)r. F. Se(_.h~fer A.J. Kru~zynski, Esq. I)r. C.J.P. de Siqueira Dr. D.C. Felton Library File No. 46D-6 AUTHORS: l).J. Wood E.B. Wilkes Copy No. I, 7 " 10 " i] " 14, 15 " 16 " 17 " lg " 19 " 2G " 21, 22 " 23 PROJECT J~B NO. 19] 2, 3,4, 5, 6 9 , 12, 13 COPY NO. t'l___

DJWIEBWIIb\/46D-6 Croup I~:?.,;c,~,rch & INvelt~pu'cnt. Ceilt-rc, Britlsh-A;:~eriean "iobacco Co. Ltd., SOUFEbVlP ION. iOth July, 1975 PROJECT WIll{AT - PART 1 CI.USTI'.'R I'IK)FILES O1" U.K. :.b\l.E SHOKERS AND "f~LEIR GENERAL SHOKING HABITS (Report No. P~.I229-R) 'i S UFIHA R Y A survey has been conducted in the U.K. among some 15OO male smokers of filter-tlpped cigarettes. The survey wa~ based on McKennel]'s questionnaire, which relates to the occasions when people smoke and the factors which motivate them to smoke. Some additional q~stions were ° asked in order to give information about the concerp felt by smokers for the possible health risks of smoking. The purpoze of the survey was to classlfy smokers into a number of categories showing distinct patterns of motivation, and different levels of so-callt r.ner Need, a," a first step towards testing the hypothesis that a smokcr'~ Inner Need level is related to his preferred nicotine delivery. This hypothesis, in turn, is seen as part of a general approach to the problem of designing cig.~zt~utc~ (~[ increa~;ed consun,er i~cceptance. The :Jns~:crs to the survey questionnaire were submitted to fackor analysis. As a result, 9 out of HcKennell's 1(1 factor:; emerged aml these included all the factors which he regarded as imported,t; a further 3 factors emerged which were not observed hy llcKennell. Clu~ter analysis J

-2- wa~ ,~erior:r.cd in order ao classify smokers into group.~ with different patterns of motivation. Twelve clusters ~:cre identified, as compared with 7 clusters in McKennell's study, These 12 clusters ware well spread along the Inner Need dimension and also along the Social dimension. In keeping with McKennell, the Inner Need score was found to be positively related to cigarette consumption, depth of inhalation and anticipated difficulty in givit~g up smoking; it was also related to the e::t=nt to which smokers supplement their regular brand by smoking plain cigarettes and hand-rolled cigarettes. Inner Need showed no relationship w~th the nicotine delivery of the filter-tipped brand usually smoked, but this could be fcr reasons of brand loyalty coupled with the rath,-r limited choice of nicotine deliveries currently offered l,v f~Itcr-tipped hran-ds in the U.K. The degree of concern for health differed considerably between clusters, but was not related to the level of Inner Need. Concern for health evidently influenced the brand choice of many different types of smoker, particularly in the direction of trying low tar brands. However in many cases their concern for health seemed to co: :t with their desire for n satisfying cigarette. The marked differences between the 12 clusters whic'a have emerged. as a result of this survey, particularly in terms of Inner Need score, suggest ci,a: el,,- ,-'lusters should form a firm b~sis for testing the hypothe.~is that Inner Need i5 related to preferred nicotine delivery. Tim testing of this hypothesis, which involves obtail:iug the reactions of 11 out of the 12 clusters to a range of experimental cign.ctLes, forms Part 2 of the present study and will be scparately reported

-3-. IN] ROI)UCT iON One of the main research objectives of Group R. & D. Centre is to put B-A.T. in a position to design cigarettes of increa,..d consumer acceptance. This objective is confined to a conmideration of ]~roduct features, and the way these influence thc consumer. Those factors which may be termed the "imagery", including brand name, pack design, advertising etc., although obviously of prime importance in r,~arketing terms, fall outside the scope of Croup R. & D. Centre. Even with this restriction the object ire is still a ve.ry broad one, and leaw:s room for a number of different approaches. The particular approach to be discussed in the present report aims to produce cert.~{n basic information, generated as a result of research among U.K. smokers, which can be of generaJ utility in a variety of markets. In considering which product features are important in terms of consumer acceptance, the nicotine delivery is one of the more obvious candidates. Others include the Caste and flavour cimracteristics of the smoke, phy~;ieal features such as draw re:;[stnnce and rate of burn, and the [,encral uniformity of the product, to n.nl;~e ' a fea'. The .importance of nicotine hardly needs to be stxesscd, ,~5 it is so widely recogniscd. Among numerous pi.er'cn of evid~nce may be mentioned an exercise conducted :;o;nc years ago hy the Imperial Tob:,cco Company. As ares "~ :: cf ::r.tensive testing among U.F. .~;mokcr~; they concluded that the oi,tiI:~,~m nicotine delivery fur the laarkct wa:; around 1.4 ,nZ per cigarette, and that .qtepw[::e reductions in delivery caused progressive rejection b2' co:l[:umer'.; (1). cill('e the al,ove rese;~r(:l, was eolnp[eted the,'. II;l~; bonn a genera[ decrease in the tar and nicotine deliveries cf tl.F< br;u,ds, V

--4 -- including the products of the Imperial Tobacco Company. The market leaders currently give nicotine deliveries of 1.2-1.3 mg, and there is a diversity of brands delivering less than i mg (2, 3). Similar reductions in delivery have been evident in other countries, notably in the U.S.A. and Cerrnany; in fact one of the ~triking features of the GeL'man market has been the growing popularity of "low nicotine" brands, krhen attempting to understand the r=ason for these trends, one is left wondering whether the present-day smoker actually prefers the smoking characteristics associated with reduced deliveries or whether he has been influenced by 'i his concern for the possible health risks of smoking and by the implication, or even the direct allegation, that cigarettes with reduced deliveries are safer. One also wonders whether the oppareut attrsction ~f reduced deliveries is true for all smokers or wh=ther it is confined Lo specific segments. These questions concerning preferred nicotine level seemed so fundamental that it was decided to try and answer them first, by direct consumer testing, before pro'cceding to consider other features of the cigarette. Yhe argument here was that, unless the (: .mer was being offered approximately the right nicotine level, it would be difficult to draw valid conclusions from tests designed to explore his reaction to different taste characteristics, different degrees of draw resistance, and so o*:. }laving decided to investigate consumers' nicotine preferences in the first instance, and allowing for the possibility that different e,~tegories of smoker might prefer different levels of nicoti':~, the findiugs o[ ~cKcunell (4, 5) seemed particularly relevant, i,~ brig, i,

-5- McKeunell co:,ductcd a sur'~.:y m:~on~ U.K. smokers, and on the basis ol t~mir answers to a questionnaire he classified them into groups with distinct patterns of smoking motivation. Some of the motivating factors, such as smoking to relieve stress, or smoking to aid concentration, appeared to represent what HcKennell termed the "inner need" to smoke, and he was able to derive a score along this dimension for each group of smokers. In a similar way he was able to derive a ~econd score along a social dimension which contrasted with the inner need dimension in that it represented the tendcncy to smok~: in variou~ social situations. He als¢~ demonstrated positive rclationahips betwecn the inner need score on the one hand and the daily cigarette consumption, the dcpth of inhalation and the anticlpated diflicull:y in giving up stroking on the other. The last-mentioned findings suggested to us that the itkner need dimension was probably defining a requirement for nic~,tine. The hypothesis was formulated that groups of smokers with a high inner need score would prefer relatively h~gh nicotine ci~,,arettes an,~ 'muld reject low nicotine ciF, arettes, whereas groups with a low inner ue., score would probably find low nicotine cigarettes quite acceptable, and might well {,refer them to those of relatively high nicotine dellvery. It wa~; ruali,':(~d, however, that these patterns of preference could easily be obscured when ~;;:u)king branded product:; because of the influx:nee of br,md luyalty, im,~,v, published smoke deliveries, etc., and than to reveal them would net'ess~tal.e conducting consumer tests w~th unbranded cigarettes in plain packs. Accordingly it was decided to administer McKmmell'.,~ questionnaire to a sample of U.K. male smokers, to repeat his factor and clusLer analysis

! f L_ -6~ assigning each respondent to a cluster, and then to tes~ the hyFochesis concerning the relationship between inner need and preferred nicer.he delivery, in addition it was decided to ask certain supplementary questions aimed at exploring respondents' concern for the health risks of smoking, because of the possibility that this might influence their brand choice and might indicate a need for specific health reassurance features in product design. The project as outlined above was given the code name ~AT; and if this project established a logical connection between the needs of the smoker and his preferred nicotine level further investigations were envisaged to encompass other importantl-produet features, with the ultimate aim of achieving a favcurable combination of attributes to suit the needs of difierent segments. It w~s believed that the principles established by this approach in the U.K. would apply, broadly, in other markets. Project ~AT is being reported in two parts. Part I, the present report, covers the administration of the McKennell questionnaire to a sample of UoK. smokers, factor analysis and cluster analysis based on their answers, details of their general sm ng habits including the brands which they currenLly smoke, and information about their concern for health. Part 2 will cover their reactions to a range of test cigarettes offering differing nicotine levels. NgTiiODS i. Tile Survey England, Crosse and Associates Ltd. were comnissioned I,y Harket Research Dept., Mkllba~k, to handle thi~; investi~:.tior,. Thuy in turn employed a computer bureau, Cybernetic~ Research Cm~a::Jtants.

r -7- °° The survey was conducted among male s::~okers only in order Lo nct)ieve comparability with McKcnncll's second stud), (5) which was based on a sample of 2,000 male smokers. A further restriction was that the product most often smoked should be a filter-tipped cigarette, since in the subsequent product-testing stage respondents tcere to be asked to smoke, and state their i, refercnce for, a range of filter-tlpped cigarettes. A sample of 1523 male smokers of tipped cigarettes were interviewed using quota sampling methods with controls in terms of four age groupings (I~-24, 25-34, 35-59, 60+) and four social class groupings (AB, CI, "i C2, DE). The quota control information was drawn from a Tobacco Research Council publication (6). Details of the sample required to meet these quota controls and the sample actually obtained are shown in A~pcndix I (page 40). Interviewing was spread throughout the country using iOO sampling points, 2 at each of 50 locatious; these locations are representative in terms of popular'ion and were selected by random methods - the full list of sampling locations is shown in Appendix II (page 41). Interviews took place in the homes of respondents and were concentrated on evenings and weekends. The questionnaire used in the survey is shown in full in Appendix III (page 42). Qx~estions 1-8 cover basic smoking in[ormation, including the number o[ tipped cigarettes smoked per day, otht:r tohace~ product:; sm "..,!, :~,, hr;~r~rl ~mnk~'d mo~t of It, n, OthtW b~'~nds :~in~kcd oc~:;l:',il na~. ly, the, d~:l,th of inhalation, etc. Q.c:~tinns 9-20 nre l.It:Fem~:li':: 42 varinbles, and it ix the answers to these which formed the ba:;~s of ~he factor an:llysis ;iml clu.':tel analysis. Question:~ 21--24 relate, tc~ the; diff[culty in givJ.g .p sl,loking.

-8- 2. Factor Aualvsls The method of deriving factors was by rotating a selected number of principal components using the well-known Varimax criteria, and then obtaining an oblique solution based upon the Varimax loadings using the • Promax method (7), HcKennell's analysis was based upon factors obtained by Vari~x rotation; the Promax method has the ability to define mote specific factors (i.e. the variance of the factor loadings is increased) but the factors so obtained are oblique. This means that the scores obtained on any one Proma× factor may be correlated to some extent with the scores on other factors. A range of factor solutions From 2 factors to 14 factors was obtained; by visual inspection the one which most closely matched McKenncll's IO factors was the 12-factor sol~tion, and this was carried forward to the next stage of cluster analysis. 3. Cluster Analysis A general description'of the cluster analysis is given in Appendix IV (page 54). The method used gave all 12 factors resulting from the factor analysis an equal opportunity of contributions to the cluster solution, and took into account the loading~ of a! f varial,}es on each of the 12 factors. This represe~ts a deliberate departure from the method adopted by McKennell (6) who only.carried forwa[d 8 of his iO factors to the clustering stage and ignored all factor loadiu~;s apart from those sho~n~ in Table i. The total sample was randomly divided into tvo halves, and cluster analysis was carried out on each half independently producing solutions ranging from 2 clusters to 12 clusters. For each solution in turn the similarity bet~een the cluster profiles in one half of the sample and those in the other half was computed by men::: o[ a

f t i i --9-- c]u,¢~cu co:n;~nrisor~ pvocram. In a;JJition [b~~ c]us:ors Zr,~%:~ c, ac)l half of . the sample were cross related to a range qf external variables such as cigarette consumption, depth of inhalation, nicotine delivery of brand most often smoked, age, social class, etc. The results of the cluster ; comparzson program indicated that the two halves of the sample matched well at both the low level (the 4 cluster solution) and the high level (the 12 cluster solution). Between these two extremes the matching became progressively less satisfactory at the 5, 6 and 7 cluster level and progressively more satisfactory at the 9, I0 and ii cluster level. After t. aking into consideration the matching data, the profiles of clusters from the various solutions and the cross relation with external variables it was agrecd that respondents should be asaigned to clusters on~the basis of a 12-cluster solution. 4. Concern for Health In order to obtain some measure of the concern felt by respondents for the harm whlci~ they might be causing to their o~n hc:alth by smoking, a battery of questions was included for this particular purpose and is reproduced in Appendix V (page 66). The cigarct, ~rands mentioned in Question 13 are those shown in the Government Clmmist's second list (2) as delivering II mg tar or less. The figure of i] mg was chosun to include Silk Cut, as it ~as felt that this brand hn~; ~1 cmlnoF;:ticm ~f "safety" '~;)' reason of tile advert, i.,;~nF, atLut:hwd to iL. [;om,- of the: questions in kppeHdi× V are taken frown ;.tcF.c.nnell'.'; fir~,t ~;tud)' (/,), others ~lere designed .special.ly for this inw'ntil;at[on. 'rm~ nnm~er in which the "health score" was derived from the ans,~ers re, que.',tim~, l,q to 2l is indicated at the end of Appendix V .... The que~;tions i,~_!.j~is i ' I . I! !