Tobacco Documents Online

. C The{uldornia Poll,}ias operated continuousl} sutee 1947 as an independent. non-partisan media spon- sored ipublic opinion news service: ThePoll is owned by Field Rescarch Corporation and since 1976 has been operated b,r The Field institil non-profit. non•partisan research groupengaged inconducting studiesof Ipubliropiniononissues of social significance. The Institute receives its financial support from ' academic.,governmental. media and private'sources: , Survey Method fntervie.vs in this survey were:made by telephone. Sample homes are dtawn iniaccordance with a probabi6tv sample design that'giNes all areas o(Ithe state and.all neiFhborhoodl aproperl3• proportion- ate chance to be included. Telephone numbers are randomly generated bti• eomputerin proportion to local prefix allocattomdensity to remove non-hstedl telcphonc biases Up to:four calls are madrweach number at different times toreach one aditlt tneach househuld An adult, respondent is selected for, the interview using an objective procedure to provtde a, balance of age andisex. Accuracy of'the Findings. Several fxuus mil be considered in assessing the accuracy of the findings in this and other Califktrnia Poll repurn One tstheamuunrof'tnler• ance in thc findings duc to the presence of random variations inhercnt in the sampling proicess itself. Anra hcr:•trc a np tnaccuracirs causrd b)^ judgcmcntal factors such as qurstHm wording and sample design. and a t fitrd are t hc cf li.•ctsRtf extcrnal events. Sampling Tibicrance Thc amount ttf sampling tnicrance in these survey findinl;n ean f><e.umauetf quite pa•ctscl; fr,• thcu:ce of x•cllitc%tinf st-rtisucal lurmulas: Thu Caltfornia Poll u.c•l an ad.ancc•d mcthal kntrvnas replicated samphng,•that prut•tdes an empirically determined estimatc of ehc range rif;sn-called sampling error for each item 4tnfhrntatKxt drveh>tted hy tha survey. This methnd wkes ammtmt of the sure of the sample. the dcFrecol vanahtht : tnresponsctocach item:,sample dl ell lclustennF: wrtgfntng). and the effects of variable tmcrvic.ver and codcr performance An estam:uc of the s'amphng,errnrranFc for this survcv is shnM•n in the table hclo.v. The sampling tolerance has been calculated at two statistueal eonlit&•ncc k•t11i. nhtch are eustumarilx used by social scienti.ts-thr y5'.: and the 9Q".'l k•.KI Too usc thr, table. first sekrt the sample svx on ..htchi the percentage in qucmwm is basedl Then note thtl plu.s and mtnux range of sampling tolerance for the dcgrcc of cnnfidence desired and apply this to the prretntaFe ligure. The, resulting "'high" and "low" estimates show the,rangc within which wecanihave; 95 : for LWh)!cuniidrn:e that if the whole popula• tionofi the:statc had bcxn1 surveyed with the same questumnaill the results of such a complrte cov. eragq would fall bctwL•eil two fiLNres obtained f'rom the data~in thetahle The sample tnicrancc figures shown in the table arc avcroFa• figures dcn.•cd fhom the actual cxperi- ence of'a number of recil survecs Tho%. represent maRlmum!tollrunccl fitr the sampll bases sho..•n.. i.e:. for xurll UtndtnEs where the dit,ision of optnHOn isaround SO"•:<i0'r1: Suran•findinE. that; sho..• a more one-sidcdldtsrnhutton of orinion. such as 70":-30.^. or aT:-10"'.. arc usuallp• subl to sliEhtly lower, sampling tolerance than those shown in, the table. Tt N-6 ~~.».. ~~• ~.~. ... ..., :. ........... . .... Plus/minus pereentage range of samplint tok•rance at - Sample Size 9Y'. conladen.e "...ronlidenor t200,....... ,. .. . 30, .............. 4.0 t000 .......... . 3.3 . . . ......... ,41. 3 f300 ............ ..717............. ,,49 600.....,....... 4;2I ............... 56 400 ............ S:Z.............. 6 9 200 ............ 7:5!............... 9 9 So............ . IS!0 .. ....... ... 19 a Other Possible Sources of Error In addition to sampling error, there are other important sources of'potcnttal inaccuracies tn thesr (and in other) poll'findings These sources include the effects of possibly biased or misleading ques. tions. possible s+•stemamc omission of'rclc.arn seg+ ments of ihr population (romthe survel samplcl and the effects of stFnifi;ant e.ents that occur during or after the time the survut tnten•ill are made There is no standard measure of these cflccts: each must be e.•al uatcd j udgmentally , Furthermore. since the influence of'these'factnrson the ultimatr,aecurac, of the sune% findtngs mal he man% times greater than thramount of sampling ernnr: it is important that they also bc carefully wetFhed So that the rradrr..•JI ha.etnformation needed to judge the possiblcimportan:e of tholl effectst ThrCalifornw Poll'prott.kl this bulletin with each relaasel drscri bi nF.thcq t.:stxmrs 1 uscd thc siza'and type of sample usod. and thedate.oftmervietsmR. The Californw Poll has an evccOcnt record for aecuraci v in reflenm} public opinion durmC its 33 year history. The stall ol Thc Cafilnrnw Poll cakes great care to formulate qlxcunns w•htch wc (cnl are objective and 1 unbiaaed and I to carefully supervise the data gathcnnF pha.~% and tnhcr rc.carch oper- ations upon tchtchl the f9r11+ findanps arc, hl Nevenhl"leSs. u1K+rs of this t;anll uther puhlic opinion polling datat shuuld!brconunu.dl, v mindful of 1a11nf ihe factnil tntltx+n.m ins toll"s aecuracv: Sampling error is not the only enterion. andi we caution against citingonl.• the sampling error fiEure trJone as the measure of i a survxl accuracv: since, to do so tends to create an,imyression ofl,aEreater degree of'precision than has in fact been achieved.. Suggested copy for editors to use when.presenting California Poll data inipublicationnr neM•scastSutwesa of~ the kind reported here by The California Poll are suhje,+t to varwhthtn duv to sampling factors and tcn other possible sources of influence on their accull The statowtdl sample results, shown in thtr rrpnn ~ are .ub- ject to a sampling tolarancr of'rlus or minus approximatcli p<•rcentagc pcomc The (ireader) (%•tewer) iGsta•nrr/ .huuld Hl.O be aware. hownver., that thirc are whar pu.+iblr sources of error for which rri:ii+e esnmatcs cannot be calculated For esamplc. different results might have hcvm uhtall trum dlllcr- ent,questinn wordinF:,and undrta-tcll tn the wac thc sampling, and tntcrn i"anF prcxc- dures wcrc carried out cxould ha- r a aFnil t;ant dftrct nrt the findings C.ollhi fxillinl: prall diminish the chances of suc,h c-rnr.l but thrv can ncac he enull ruled out Ia is af.n pu.. siblc. oflcoursef that rventc tkcunnE .mvr the timc the interaiea•s were eonditetcd could hl chan(;ed'the opinions reported here ( (