THE POLLS-A REVIEW SURVEY ORGANIZATION DIFFERENCES IN ESTIMATING PUBLIC PARTICIPATION IN THE ARTS JOHN P. ROBINSON

Transcription

1 THE POLLS-A REVIEW SURVEY ORGANIZATION DIFFERENCES IN ESTIMATING PUBLIC PARTICIPATION IN THE ARTS JOHN P. ROBINSON There are two main sources of arts participation data for the United States. One is the survey of public participation in the arts (SPPA '82 and SPPA '85), conducted for the National Endowment for the Arts by the U.S. Bureau of the Census across the full twelve months of 1982 and the first six months of In each month of those years, a separate random sample of about 1,500-2,000 individuals aged 18 and over were interviewed personally as a supplement to the National Crime Survey. The SPPA's "core" questions concerned the extent to which respondents had attended arts performances (e.g., jazz concerts, opera, ballet) in the previous year. In March 1984 (as in 1980 and 1987), Louis Harris and Associates (1984) independently conducted its "America and the Arts" survey. This public opinion study included several questions on the public's participation in these same types of arts activities. While the specific questions did vary from those asked in the SPPA and there are no outside validity data for complete verification, there is enough commonality in content and consistency in results across activities to isolate significant differences across the two survey organizations. This review compares the survey estimate of arts and recreational participation for these two major data sources-one governmental and the other commercial. The self-report questions are not identical and differ somewhat in time of data collection so that the comparisons are not true experiments in the sense that the classic Stouffer (1955), Steiner (1963), Schuman and Presser (1981), and Bishop, Oldendick, and Tuchfarber (e.g., 1982) studies were. The analysis does involve comparing some questions that involve multiple behaviors in the same item (e.g., "Did you do activity x or y?") with those that ask about a single activity ("Did you do activity x?" and "Did you do activity JOHN P. ROBINSON is Professor of Sociology at the University of Maryland, College Park. He wishes to thank Tim Triplett and Sue Dowden for their field expertise in designing and executing the experiments described in this article and also to thank Stanley Presser, Seymour Sudman, and anonymous reviewers for their advice and suggestions-not all of which were gracefully received or followed. Public Opinion Quarterly Volume 53: ? 1989 by the American Association for Public Opinion Research Published by The University of Chicago Press / X/89/ /$2.50

2 398 John P. Robinson y?"); these questions may not be equivalent in terms of respondent perception or response. Table 1 describes the basic parameters of the 1982 and 1984 studies, which were designed similarly and produced findings consistent with those of other years for both survey organizations. It can be seen first that SPPA '82 had over ten times the sample size (N = 17,254) of the Harris study (N = 1,504); in 1985, identical SPPA procedures were used with 13,675 respondents across the first six months of the year. SPPA '82 was conducted across the full 12 months of the year, rather than the single month of March covered in the Harris survey. (The March survey did not produce different results from the eleven other months in SPPA '82.) Most of the SPPA data were collected by inperson interviews conducted in the respondents' homes; interviews were conducted by telephone in the Harris study. The sampling frame for SPPA included all households in the country; as in 1980 and 1987, the Harris sample frame was by nature restricted to households with telephones. In the Harris study, one individual aged 18 or older within each selected household was chosen by the oldestyoungest, male-female technique; in SPPA, all individuals aged 18 or older were selected into the sample. The Harris study reports no overall response rate figures. Two callbacks were made over a three-day period to nonresponding numbers, and one attempt was made to convert refusals. The Harris (1984) methodological report does not give figures on how many telephone numbers were dialed or how many refusals were encountered. In SPPA '82, the Census Bureau achieved a response rate of over 85% for every month; this is lower than most Census Bureau studies, mainly because it was the "exit interview" from the National Crime Survey, and most respondents had been interviewed six times previously over a threeyear period. The last column of Table 1 provides parallel details for a third national survey, one conducted by the University of Maryland's Survey Research Center. The ARTS '83-84 survey was designed as an expansion of the SPPA '82 survey, but included many of the same core questions as its initial items. The sample, however, was a completely independent one, conducted by RDD telephone (like the Harris Surveys) but in two separate waves-one in June 1983 and the other in January Internal Factors Affecting Response Differences The effects of four "internal" factors (context effects, question time frame, question wording, and data weighting) are assessed in the analysis which follows. The effects of "external" factors (e.g., sample frame, response rates) are discussed in the concluding section of this article.

3 Table 1. Differences in Basic Survey Samples and Procedures Harris 1984 SPPA '82 ARTS '83-84 Data collection Louis Harris and U.S. Bureau of Survey Research agency Associates, Inc. the Census Center, University of Maryland Sample size 1,504 17,254 1,054 Survey dates 5-15 March Jan-30 Dec 1-30 June 1983 (using same 1982 (same 2-24 Jan 1984 questions asked questions used in 1980 and in SPPA '85 and 1987 surveys) ARTS '83-84) Survey method 100% telephone 75% personal, 100% telephone 25% telephone Sample frame Random digit dial Based on 1980 Random digit dial (response (unknown) Census (68%) rate) (85%-95%) Respondent One person aged All residents of One person aged selection 18 + per house- households 18 + per house- (within sam- hold (of those aged 18 + hold (with ple house- home at the call-backs to hold) time of the call) designated respondents) Prior inter- None (first con- Six (over preced- None (first conviews tact) ing three years) tact) Data weighting By age, sex, and By age, sex, and By age, sex, and race race education Preceding ques- Amount of leisure None (lead ques- None (lead tions time tions)-al- questions)- Access to leisure though earlier although earlier facilities questions had questions had Importance of asked about ex- asked about museums, the- perience with news informaaters to busi- crime tion and news ness and econ- media use omy Should be more or less (sports or arts) events given in the area Importance of creative activities Reasons for not going to arts activities Do arts activities (see Table 3)

4 400 John P. Robinson CONTEXT EFFECTS For several years, the prevailing wisdom in the survey research community has been that it makes little difference where questions were placed in a survey questionnaire or interview. More recently, several studies have indicated that preceding questions can sometimes seriously affect the responses that people give to later questions, particularly if some expectation or mindset is built into these preceding survey questions. In the case of arts-related behavior, for example, prior questions that imply that respondents ought to be participating in such behavior, that other people are participating, or that identify the purpose (or the sponsoring agency) of the study could well skew responses to the survey questions away from responses that would be obtained if no such prior questions were asked. In the case of the Harris survey, no blatant context effects appear to be present, although that remains an empirical question. There are no introductory statements about the purpose of the survey or the sponsoring agency (Phillip Morris Company). However, as noted at the bottom of Table 1, there were several questions about the arts and leisure that preceded the arts participation questions-questions about the importance of the arts to the economy and the importance of creative activities generally to the respondent, or about whether the respondents themselves had engaged in various arts-related activities. There is the possibility, then, that these questions could create the impression that reporting more attendance at arts performances is what is expected of respondents in the survey. Respondents might want to report more arts activity, say, as a way of compensating for not being active in the arts in the prior year, or as a way of being consistent with their earlier statements of support for the arts (e.g., 83% of Harris survey respondents said it was at least somewhat important for them to have more creative activities in their community). Alternatively, one could argue that these warm-up questions served to refresh respondents' memories regarding arts participation. In the case of SPPA, such a source of prior question bias is not possible since these were the first questions asked of respondents in this supplement to the National Crime Survey. However, the sponsoring agency of the survey was clearly identified to respondents prior to that first question: "The Census Bureau is collecting this information for the National Endowment for the Arts." This by itself could be a source of higher reporting since respondents knew to which organization to give possibly compliant or helpful responses. Whether this would encourage higher responses than the Harris lead-in questions is not clear, although results from the ARTS '83-84 study that did not mention NEA sponsorship suggest the SPPA responses were not affected by the introduction.

5 Poll Review: House Estimates of Participation Rates 401 Table 2. Percentages in the 1984 Harris Surveys Reporting Attendance Ever and Last Year Ever Attended Harris Activity Attended Last Year Movies NA 78 Live performance of plays, musical comedies, pantomime, other theater Live popular music performances by popular singers, bands, rock groups Visits to art museums that exhibit paintings, drawings, sculpture NA 58 Live performances of opera or musical theater Live performances of ballet, modern dance, folk or ethnic dance Live performances of classical or symphonic music by orchestra, chamber groups, soloists NOTE: NA = question not asked. QUESTION TIME FRAMES As shown in the Appendix, the Harris survey asked several of its attendance questions using a two-stage filtered approach. It first asked whether the respondent ever went to any such performances; for each "yes" response it asked how many times the respondent had attended in the previous year. The SPPA question asked directly whether the respondent had gone in the last year. No solid survey evidence seems available to show that asking a longrange filter question will affect response to a follow-up shorter-range question. However, it would not be surprising to find that having already said that one had engaged in a (socially desirable) activity increased the likelihood of also saying one has done that activity in the shorter run. Therefore, however well-intentioned the filter is in sparing respondents the burden of describing an activity in which they do not engage, responses to the second-stage questions could well be affected by the process of being filtered from the long-run question. Respondents may then feel inconsistent if they report not having done the activity in the short run. Thus, at least 80% of Harris respondents who said they had ever done the activity said they had also done it in the last year. As shown in Table 2, the "ever attended" Harris figures are only 5 to

6 402 John P. Robinson 7 percentage points higher than the Harris proportions reporting attendance in the last year. We shall return to this point in the conclusion. QUESTION-WORDING DIFFERENCES There are some significant differences in the question wording of the two surveys, particularly regarding which activities are combined in the question definition. As detailed in the Appendix, we find the following contrasts in activity definitions and combinations: Harris 1984 SPPA '82 1. Plays and musicals, panto- 4. Musicals and operettas mime, and other theater 5. Nonmusical stage plays 2. Art museums (with exam- 7. Art museums and galleries ples) 3. Opera or musical theater 3. Opera 4. Musicals and operetta 4. Ballet, modern dance, 6. Ballet folk/ethnic Modern (ARTS '83-84) Folk/ethnic (ARTS '83-84) 5. Classical music/symphony 2. Classical music (includes choral music) The Harris survey question combines plays and musicals in one question, and combines opera and musical theater (thus allowing attendance at a musical the opportunity to be counted in two separate questions). The Harris dance question also combines ballet, modern dance, and folk/ethnic dance into a single question; the three were asked separately in the ARTS '83-84 survey. The SPPA survey question, on the other hand, may produce higher estimates on musicals and classical music items because they explicitly include operetta with musical theater and include choral music with classical concerts. One way of partially resolving these question discrepancies is to total the figures for the two separate SPPA items to make them equivalent to the Harris combined item-at least conceptually. Thus, we combine the SPPA items on plays and musicals and on opera and musical theatre. For the Harris item on ballet, modern dance, and folk/ ethnic dance, we can accomplish this goal only through the separate questions in the ARTS '83-84 study that were included on modern dance and on folk/ethnic dance performances. The result of all these combinations is shown in Table 3. It can be seen that even with these recalibrations, the Harris and SPPA surveys diverge widely and in a quite consistent direction: the Harris data

7 Poll Review: House Estimates of Participation Rates 403 indicate far higher participation. Thus 56%-60% of Harris's respondents report theater attendance in the previous year, compared to 21%-23% of SPPA respondents who reported that in the previous 12 months they had either attended a musical or a nonmusical play (and subtracting out the proportion who did both and would otherwise be doubly counted). In the case of museum visits, the Harris data show 55%-60%, which is almost twice as high as the 22%-28% in the SPPA surveys. For dance, the Harris figures are 20%-28% and the ARTS '83-84 figure is 13%; for classical concerts-the most directly comparable item-we find Harris at 21%-29% and SPPA at 13%. In the closest set of figures, 20%-28% of Harris respondents reported attending opera or musical theater, compared to the equivalent combined figures for 18%-20% in SPPA. With this exception, all differences are well beyond the.001 statistical significance level. In contrast to the 2- to 5- point sampling error levels theoretically associated with such surveys, the last column of Table 3 shows discrepancies of up to 36 percentage points, 19 points on average. These higher differences in reported participation levels in the Harris data are not confined to performance attendance data. As shown in Table 4, the Harris figures for self-participation in various art forms are also markedly higher. It should be noted that there are even more serious question-equivalence problems with these self-performance activities than with the attendance data. Moreover, the Harris data use a broader time frame, including activities done "at least every once in a while," which would probably be perceived by most respondents as extending beyond the one-year time frame used in the SPPA questions. (At the same time, the differences in the two columns of Table 2 suggest that the yearly and lifetime time frames should only differ by about 5 to 7 percentage points.) In summary, it would appear that question wording and definition per se also do not seem to account for the large differences in reporting levels in Tables 3 and 4. The question filter may have had some effect in the higher Harris figures; this explanation will be examined more directly below. We return now to a factor that should produce more of a difference. DATA WEIGHTING Both the Harris and SPPA '82 were weighted to reflect population totals. The Census Bureau weights reflected the 1982 Census population counts; the population frame of reference employed by the Harris data is less clear. However, Table 5 shows the two samples do not diverge greatly on the factors for which the Harris data have been weighted, namely, age, sex, and race; no difference is greater than 3

8 oo??oo 00 g m t N g ^? N~r = e? N z N ; b g~~~~~~~~~4 404 ot?

9 00 (7N (7N "D if) if) if) if) if) if) cn o 4-o 0 C) 0 42 v) cn -4 4-i

10 'E 4--4 CZ 00 an Z Z Z Z CZ 00 4-j 1 C r- r-...4 m Z Z Z z > 00 4-j CZ C IC z 00.0 Z Z Z 00 C r-,t,t an,t rn 00 an W') rn ON an C "tt rn cd 4-, > cd cd cd 64 cd 45 C's cd m cn IZ u iz ct cd Ct cd C), '(: 4) U) 4-, tio i 4. cd cd 406

11 Table S. Demographic Composition of the Sample Harris 1984 SPPA '82 Number of Weighted Weighted Interviews % of Total % of Total Nationwide 1, % 100% Region East Midwest South West Size of place Cities Suburbs Town/rural Age Education 8th grade th grades not reported 13 High school graduate Some college College graduate Sex Male Female Race White 1, Black Hispanic Income $7,500 or less $7,501-$15, $15,001-$25, $25,001-$35, l 25 $35,001-$50, J $50,001 and over

12 408 John P. Robinson percentage points. The two samples are also relatively close on the factors of region, urbanicity, and, generally, income. The SPPA data were self-weighting by household size, since all adults were interviewed; the Harris data were apparently not weighted by household size. However, for level of education, the factor that the SPPA '82 data clearly show makes the most difference in arts attendance, there is a serious departure. It is reported that the Harris survey interviewed only 49 respondents with less than a high school degree, which Harris reports as representing only 4% (actually 3.3%) of this sample. Yet the Census Bureau puts the percentage with only a grade school degree at 12% and another 13% with 9-11 years of education; both groups combined, then, constitute more than a quarter of the American population. With less-educated respondents being so seriously underrepresented, it should not be surprising to find the Harris arts participation figures being much higher.' Nonetheless, even this substantial discrepancy still does not account for much of the differences in results. Table 6 arrays the educational level differences within each survey, thus directly comparing the participation rates of grade school educated, college educated, etc., across each survey. While the comparison is not exact because the Harris data are for the "ever attend" responses rather than for attending in the last year, we have seen in Table 2 that these differences between Harris ever attend vs. yearly estimates are not great (being 5 to 7 percentage points). Yet, as Table 5 shows, the Harris data estimates continue to be double, or up to 46 percentage points higher (for visiting art museums among the "some college" group) than the rates reported in SPPA '82 for equivalent educational categories.2 It should be noted that properly weighting the Harris results by 1982 Census Bureau education data does reduce Harris estimates in the expected downward direction; for example, after proper weighting for education, Harris's classical music attendance moves from 34% to 28% and movie attendance from 78% to 67%. Nonetheless, while the properly weighted Harris data do come closer to the SPPA '82 data, they remain significantly different even after these factors are taken into account. The same deflation would likely be found if the Harris data could be 1. Data from Presser, 1981, indicate a continuing long-term tendency for Harris data to have disproportionately higher numbers of college-educated respondents compared to Census Bureau figures. 2. That is not to say that the relationships are identical, and a log-linear analysis on the theater data in Table 6 (performed by a reviewer) shows that there is a significant threeway interaction indicating a different relation with education in the two surveys.

13 Poll Review: House Estimates of Participation Rates 409 Table 6. Differences in Attendance by Education Levels: Harris 1984 and SPPA '82 High School Some College 8th Grade Graduate College Graduate Theater Harris % 54% 77% 88% SPPA ' Art museums Harris SPPA ' Opera, musical theater Harris SPPA ' Classical concerts Harris SPPA ' NOTE: Harris 1984 question asks "ever attended"; SPPA '82 asks "attended last year." weighted for nontelephone ownership. However, after control for education, race, urbanicity, and sex, Robinson and Triplett (1987) found telephone owners in the SPPA reported only one- to two-percentagepoint higher participation than nonowners for most arts activities. Thus, weighting by the above demographic factors accounts for most of the differences by telephone access. Conclusions We have examined several "internal" factors that may explain the far higher arts participation rate figures in Harris data than Census Bureau studies. Four of these factors that were reviewed did seem to account for some of the discrepancy: 1. The use of the "ever go" filter rather than the direct "last 12 months" time frame in the initial Harris question 2. The inclusion of multiple activities (opera and musicals; ballet and modern dance and folk/ethnic dance) in some of the Harris questions 3. The underrepresentation of respondents with less than a high school education in the Harris sample-as well as the failure

14 410 John P. Robinson generally to weight the Harris data by the proportion of respondents in various educational categories in the population 4. The exclusion of nontelephone households from the Harris sample frame The possibility that the Harris figures might be higher because of question context effects was also raised, although there were reasons to expect SPPA responses to be inflated for that reason also. The activity definitions in the wording of the questions did not otherwise appear to be a major consistent source of higher reporting levels, although this does remain a potential source of upward reporting. Thus, while a good portion of the variance in results can be explained by these factors, much of it cannot be. The four factors noted above each account for higher reporting levels by up to a fifth, while the comparisons in Tables 3 and 4 are sometimes doubled. As Table 6 shows, the proportions in the Harris survey continue to be markedly higher for the same educational levels of respondents. Thus, some other major factor must be at work to produce the discrepant figures. One way of identifying such a factor is through separate experimental studies using the same Harris and SPPA questions. Two such smallscale experiments were conducted by the University of Maryland Survey Research Center, one in spring 1987 and the other in fall 1987 with statewide RDD samples (response rates 70% and 74%). These omnibus Maryland Polls asked half the sample the Harris question and the other half the SPPA question. These questions were inserted roughly midway through these omnibus surveys with unrelated questions preceding them. For the most comparable item, namely, attending classical music concerts, the population estimate for participation in the previous year was about the same: 32% for the SPPA question (#2; N = 243) and 34% for the two-stage Harris question (#10c-d; N = 253). For attending art museums, the single-stage Harris question (#13a; N = 262) produced a 47% estimate, the SPPA question (#71; N = 303), 44%. Sample sizes (about 250) in each condition were rather small in relation to those in Tables 3 and 4, but the results were well within sampling error. This would seem to rule out question filter effects as a major explanation. Thus, we are left with external factor explanations such as cooperation rates and respondent selection procedures. As noted earlier, we expect that while no response-rate data are provided in the 1984 Harris report, standards on such methodological matters tend to be more relaxed in commercial surveys. Surveys done with careful probability methods and higher response rates are more expensive to conduct, and minimization of costs is of paramount concern in most commercial

15 Poll Review: House Estimates of Participation Rates 411 surveys. Moreover, the results from telephone and personal interviews usually can be made to match population quotas well; and it may be argued that, if people who respond to these surveys seem little different from those who refuse or are hard to reach, the extra cost for higher response rates would not be justified. In this case, however, the behavior under study appears to be much more sensitive to relaxation in respondent selection procedures. Possibly respondents who take part in these surveys are more interested in the arts than those who refuse or are hard to reach. For now, that selectivity factor appears to be the most plausible explanation we have for these data divergences on arts and cultural participation. With their longer history and their ability to pay more careful attention to methodological detail, one might expect the Census Bureau figures to be a more trustworthy source than commercial survey data. That seems to have been the case for the 1977 National Recreation Survey, for example (Robinson, 1986). At the same time, other studies done by commercial survey firms often do produce results consistent with those from Census Bureau or careful academic surveys. That does not mean that the Harris or other commercial data do not provide valid insights into the correlates of participation. Indeed, their relations of participation to demographic background factors (e.g., education, age) are generally similar to those from SPPA and other Census studies. It may also be the case that the Harris (1987) data can provide valid perspectives on trends in participation (e.g., both show declines in musical theater attendance across the mid 1980s). Nonetheless, it is clear that the Harris data do provide markedly higher estimates of participation than those asked by the U.S. Census Bureau. Without validity data on the actual participation of the American public (which is all but impossible given the annual time frames involved) or data from a true controlled experiment across survey organizations, it is hard to tell definitively whether the Census Bureau data are too conservative or the commercial data are too generous. Our small-scale studies again suggest the latter conclusion, but they need replication and validation. These results also challenge the conventional conclusion that "house effects" across survey organizations are minor and unimportant, as first reported by Stouffer (1955) and essentially replicated in Steiner (1963), Smith (1982), and Turner and Martin (1984). At a time when it is proposed that commercial firms can collect survey data more cost effectively than government agencies, the present findings have important policy implications for which types of agencies should collect behavioral and participation data.

16 412 John P. Robinson Appendix: Arts Participation Questions Employed in the Harris and SPPA Surveys ATTENDANCE ITEMS Harris SPPA '82 and '85 (1980, 1984, 1987) (+ ARTS '83-84) 7. Do you ever go to any live per- 5. (During the last 12 months), did formances of plays, musical com- you go to a live performance of a edies, pantomime, or other kinds nonmusical stage play? Do not of theater or not? include grade school or high school productions. 8b. (If do go) Approximately how 4. During the last 12 months, did many times did you go to live you go to live musical stage play theater performances in the past or an operetta? Do not include 12 months, not counting any per- grade school or high school proformances given in connection ductions. with schools or classes? 9. Do you ever go to live popular ARTS Did you go to any other performusic performances by popular ' mances of live music? (Asked afsingers, bands, or rock groups ter questions about jazz, classical, and the like? opera, and music theater) loc. What about live performances of 2. During the last 12 months, did classical or symphony music by you go to a live classical music orchestra, chamber groups, solo- performance? This includes ists, etc.-do you ever go to music and instrumental or vocal such classical music perfor- recitals, as well as symphony and mances or not? choral performances. lod. (If do go) Approximately how many times did you go to live classical music performances in the past 12 months, not counting performances given by your children in connection with schools or classes? loe. What about live performances of 3. During the last 12 months, did opera or musical theater-do you go to a live opera? you ever go to live opera or mu- 4. During the last 12 months, did sical theater performances, or t.. not? you go to a live musical stage play or an operetta? Do not inlof. (If do go) Approximately how clude grade school or high school may times did you go to live per- productions. formances of opera or musical theater in the past 12 months, not counting performances given by your children in connection with school or classes?

17 Poll Review: House Estimates of Participation Rates la. And what about dance perfor- 6. During the last 12 months, did mances-do you ever go to live you go to a live ballet perforperformances of ballet or modern mance? dance, or folk or ethnic dance, ARTS In the last 12 months, did you go or not? ' to any other type of live dance 1 lb. Approximately how many times performance, for example, moddid you go to live performances ern dance, ethnic or folk dance, of ballet or modem dance, or jazz dance, or tap dance? folk or ethnic dance, in the past 12 months, not counting performances given by your children in connection with school or classes? 13a. How many times, if any, did you 7. During the last 12 months, did visit art museums that exhibit you visit an art gallery or an art paintings, drawings, sculpture, museum? etc., during the past 12 months? 13b. And how many times, if any, did 24. During the last 12 months, did you visit science or natural his- you visit a science museum, natutory museums during the last 12 ral history museum, or the like? months? 13c. And how about history museums 25. During the last 12 months, did which preserve objects from the you visit a historic park or monupast-including historic buildings ment, or tour buildings, or neighor sites-how many times did borhoods for their historic or deyou visit history museums in the sign value? past 12 months? SELF-PARTICIPATION ITEMS 6. Let me read you some activities During the last 12 months that some people do at least every once in a while. Please tell me whether you yourself do each of these activities at least every once in a while, or not Paint, draw, or engage in graphic 35. Did you do any painting, drawarts such as etching ing, sculpture, or printmaking activities? 6.2. Make pottery or ceramics 29. Did you work with pottery, ceramics, jewelry, or do any leatherwork, metalwork, or similar crafts? 6.3. Sing in a choir or other choral ARTS Have you take singing lessons or group ' done any singing for your own pleasure? Have you sung in a choir?

18 414 John P. Robinson 6.4. Do needlepoint, sewing, or other 30. Did you do any weaving, handwork crocheting, quilting, needlepoint, sewing or similar crafts? 6.5. Write stories or poems 33. Did you work on any creative writings such as stories, poems, plays, and the like? Exclude any writing done as part of a course requirement Play a musical instrument ARTS Have you taken music lessons or ' played musical instruments for your own pleasure? 6.7. Dance ballet or modem dance ARTS Have you taken any dance ' classes or done any dancing for your own pleasure-ballet and modem dance? 6.8. Dance folk or ethnic dance -Folk and ethnic dance? References Bishop, George, Robert Oldendick, and Alfred Tuchfarber (1982) "Effects of presenting one versus two sides of an issue in survey questions." Public Opinion Quarterly 46: Louis Harris and Associates (1984) Americans and the Arts. New York: Louis Harris and Associates. (1988) Americans and the Arts. New York: Louis Harris and Associates. Presser, Stanley (1981) "The Harris Data Center, Harris national surveys, and trend analysis." In E. Martin, D. McDuffee, and S. Presser (eds.), Sourcebook of Harris National Surveys. IRSS Technical Papers, no. 6. Chapel Hill: Institute for Research in Social Science, University of North Carolina. Robinson, John (1986) American Recreational Trends, College Park, MD: Survey Research Center, University of Maryland. Robinson, John, and Timothy Triplett (1987) "Activity pattern differences between telephone and non-telephone households." Paper presented at the Conference on Telephone Survey Methodology, Charlotte, NC, November. Schuman, Howard, and Stanley Presser (1981) Questions and Answers in Attitude Surveys. New York: Academic Press. Smith, Tom (1982) "House effects and the reproducibility of survey measurements." Public Opinion Quarterly 46: Steiner, Gary (1963) The People Look at Television. New York: Knopf. Stouffer, Samuel (1955) Communism, Conformity, and Civil Liberties. Reprint. Gloucester, MA: Peter Smith. Turner, Charles, and Elissa Krauss (1978) "Fallible indicators of the subjective state of the nation." American Psychologist 33: Turner, Charles, and Elizabeth Martin (1984) Surveying Subjective Phenomena. New York: Russell Sage Foundation.