Creating a Major National Longitudinal Cross-sectional ...



Creating a Major National Longitudinal Cross-sectional Database: The National Science Foundation Surveys of Public Understanding of Science and Technology

Susan Carol Losh

Department of Educational Psychology and Learning Systems

Florida State University

Tallahassee FL 32306-4453

slosh@fsu.edu

Paper to be presented at the Annual Meetings of the American Educational Research Association, Chicago, April 9-13 2007.

This research has been supported by an American Statistical Association/National Science Foundation Research Fellowships, and grants from the American Educational Research Association and the National Science Foundation to Susan Carol Losh; none of these agencies bear any responsibility for the findings or my interpretation of findings. My thanks for insights over the years to Ryan Wilke and Michael Quinn, Alice Robbin, Ray Eve and Martin Bauer.

The construction, dissemination and use of two longitudinal archives created from the National Science Foundation Surveys of Public Understanding of Science and Technology is described. The data are the most comprehensive compilation of science knowledge and attitudes among the American adult general public and provide the most educational detail of any large-scale survey of U.S. adults. One archive tracks information technology access and use (1983 to 2002). The other has nearly 300 variables (n ~22,000) on science interest, factual knowledge, inquiry understanding, pseudoscience beliefs, general attitudes and attitudes about science policy (1979 to 2003). The data are now online at several sites and can be analyzed online or downloaded, thus expanding their venue to the entire scholarly community.

Although with the creation of the Statistical Abstract of the United States, U.S. federal government compilations of social indicators now stretch back at least 125 years, more specialized collections of indicators are much younger. For example, the FBI Uniform Crime Reports debuted in 1929, the Digest of Educational Statistics was mandated in 1962, and the Integrated Post-Secondary Education Data System (“IPEDS”) began in 1992. “Social indicators” research gathered considerable momentum during the late 1960s and early 1970s, when, in addition to assembling data on behavior or achievements (e.g., counts of murders, numbers of high school and college degrees, or labor force participation), researchers launched national knowledge, opinion and attitude surveys ultimately designed to be repeated over time. A notable early example is the General Social Survey (GSS), which completed its 34th year of national probability samples of in-person surveys of adults in 2006.

The creation of these longitudinal databases has enabled several types of research. First are genuine panel studies in which the same individuals or organizations are repeatedly studied, such as IPEDS. In the Census Bureau’s Survey of Income and Program Participation (“SIPP”), national samples of participants rotate in and out of the study. More common are repeated cross-sectional surveys, in which different probability samples of respondents are asked the same sets of questions over time, making “gross tracking” possible (e.g., the GSS). Repeated cross-sectional studies then make possible the creation of synthetic cohorts, which can be, with caution, followed over time to compare and contrast issues of aging and generation effects.

This presentation discusses my creation and use of longitudinal cross-sectional archives from the National Science Foundation Surveys of Public Understanding of Science and Technology. The larger archive currently comprises about 300 variables from nearly 23,000 survey respondents, spanning 1979 to 2003; it is the most comprehensive study of American adult public understanding of science. These surveys also contain also more detail on adult high school and college educational experiences than in any other U.S. adult survey. The first part of my presentation describes the basic NSF Surveys data, which provided the input for these archives.

I created two archives from the original data: one on information technology access and use (1983-2002), which has the longest known American timeline on IT, and a second, much larger archive, containing the core material on science knowledge and attitudes (1979-2003). Both have extensive demographic information as well; both are now available online and can be analyzed online. The IT archive is available at (1) the University of Maryland’s WebUse site and on (2) the University of Michigan’s Inter-University Consortium for Political and Social Research (ICPSR) site. The core science knowledge and attitudes archive is available on (1) the ICPSR site and on (2) the University of Connecticut’s Roper Center for Public Opinion Research online archive collection. Part of my presentation addresses the “politics” of placing a data archive online with a major archival site, such as ICPSR or the Roper Center, and questions that I wish I had asked before doing so.

Even before I placed the archives online, I had responded to several requests from American and international scholars who wished to analyze the data. As co-Principal Investigators until 1999, Drs. Jon Miller and Linda Kimmel, now at Michigan State University, most often have used these data in research (e.g., Miller, 1983; 1998; 2004; Miller & Kimmel, 1998). Miller was the original PI on the NSF Surveys, directing the project from 1979 to 1999 for 20 years; he has worked with Dr. Kimmel for over two decades. However, I have now also extensively analyzed the data and published the results (e.g., Losh, Tavani, Njoroge, Wilke & McAuley, 2003; Losh, 2003; 2004a; 2004b; 2006a), and the data are being used for international comparisons with the Eurobarometer and other surveys (e.g., Allum, Sturgis, Tabourazi & Brunton-Smith, in press; Bauer, Allum, & Miller, 2007). The final portion of my presentation gives examples of some of these analyses.

AMERICAN CIVIC SCIENCE LITERACY AND THE NSF ARCHIVAL DATA

The U.S. general public is enamored of science and technology, and U.S. expenditures on science and technology research and development in 2003 alone reached 284 billion dollars (U.S. Bureau of the Census, 2006, Table 774). However, although Americans express sizeable interest in medicine, science, and technology, numerous recent controversies, e.g., biotechnology applications, require an understanding of basic science that many adults may lack. Currently, a plethora of American journalists, politicians, social and behavioral scientists, and educators assert that our youth are unprepared for college science, that students switch from science concentrations, and adults cannot discuss science at the level of a major newspaper. When we juxtapose spending cuts on science education research against national and international needs for trained personnel, these concerns become more urgent (Burris, 2006; Miller, 2000, Lemonick, Keegan, & Ybarra, 2006; Seymour, 2006, Weiman, 2006).

Public Understanding of Science (“PUS”) or Civic science literacy (“CSL”; see Bauer, et al., 2007) among adults in particular is often critical for intelligent policy discussions and a supportive research climate (Allum, et al., 2007, in press). Individuals who possess more accurate science and technology information can rely less often upon self-proclaimed “experts” and thus may be able to partly defuse purely political influences on policy. Some scholars (e.g., Miller, 2007) argue that high CSL can even enable adults to withstand the frequent “assaults” from pseudoscience peddlers in a variety of media.

In addition, understanding how science “works” enhances the evaluation of information from school, work, friends, and cultural channels, such as media, and CSL among adults can influence their input into schools and other cultural institutions. Scientifically literate citizens can enjoy an enhanced quality of life: better health practices, more enjoyment of media, entertainment, or sports consumption, and greater ease of adaptation to new technology. In turn, the use of new technology, such as the Internet, can facilitate knowledge.

Indicators of general public science and technology knowledge and attitudes can provide important information; for example, some data may provide a kind of “national report card” for adults. In one recent analysis, Losh (2006a) found that, contrary to beliefs among many science college faculty (e.g., Weiman, 2006), more recent generations of American adults have higher CSL levels than earlier cohorts, even when educational variables, such as achieved degree level, or high school or college science exposure, are controlled. This is one illustration where we can begin to compare age versus cohort effects through building synthetic generations from a longitudinal, cross-sectional database (also see Losh, 2007).

We can also study whether relationships between background characteristics or science knowledge and attitudes, with other variables controlled, e.g., dimensions of education, remain constant or change over time. For example, in another recent analysis (Losh, 2006b), I found that gender had a somewhat greater impact in how positively adults felt about science as a career choice for themselves or their children in 2001 than in 1983.

American surveys about science and technology adult “literacy” date from at least the 1950s; however, the most extensive studies of the U.S. adult general public are the 1979-2003 NSF Surveys of Public Understanding of Science and Technology (“the NSF Surveys”, directed by Jon D. Miller 1979-1999, ORC-MACRO 2001, and the Joint Program in Survey Methodology, Universities of Michigan and Maryland). The 2006 data were collected through the GSS and NORC-Chicago. The Surveys coordinate with several international surveys, such as the “Eurobarometer” (e.g., Allum, et al., in press; Bauer, et al., 1994). Except for 1979, which consists of in-person surveys, the NSF Surveys use Random Digit Dial telephone interviews in 12 separate years (1979. 1981, 1983, 1985, 1988, 1990, 1992, 1995, 1997, 1999, 2001, 2003). The 2006 GSS data are in-person surveys. In addition, the GSS employed a split-ballot (randomized two condition experimental) mode, in which about 1500 respondents received many of the earlier trend data questions, while the other half received new questions, including items about the cultural authority of science and about Polar Regions.

Response rates vary on the Surveys, and, comparable to other telephone surveys, dropped during the 1990s. The 2001 survey had a 51 percent “hit rate” (percentage completed when a human contact was made) and the 2003 had a 54 percent “hit rate”. The General Social Survey is typically a bellwether for other national studies, and its response rates in eligible households have recently been in the low 60s.

Items on the NSF Surveys monitor dimensions of science or technology interest, knowledge, and attitudes as well as cultural consumption. Questions include science factual knowledge (the so-called “Oxford items”) understanding science inquiry, positive and negative attitudes toward science, stereotypes about scientists, “traditional” (e.g., astrology) and “modern” (e.g., space aliens) pseudoscience beliefs, and attitudes about several dimensions of science-related policy (e.g., global warming, genetically engineered foods).

The NSF Surveys contain many standard demographic variables (e.g., marital, parental and labor force status; ethnicity), and they also include considerable detail about educational variables. In addition to degree attainment, there are data about high school exposure to math, biology, chemistry and physics, liberal arts versus professional preparatory college training, college major, and the number of college science courses the respondent elected. The data about high school experiences are particularly pertinent because only one-third of the U.S. adult population has at least an Associate of Arts degree.

If one wishes to assess issues in science and technology cognition over time, which I do, it was clear that a longitudinal, repeated cross-sectional archive had to be built from the separate NSF Surveys files. Such an archive is a necessity to assess the effects of time, and in my own and other analyses (e.g., Losh, 2006a; 2007; Miller, 2007) the effects of birth cohort or “generation”. Further, once built, cross time and cohort analyses can be quickly and easily done, tests of statistical significance over time easily calculated, and a foundation created to add further years as the NSF Surveys series continues.

CREATING THE ARCHIVES

There have been several prior or concurrent attempts to create a longitudinal national database of the NSF Surveys. Most, however, have been abandoned or stalled. A probable major reason for these delays and false starts is that there is just so much data in the original files; so many variables over the 20-odd years of the separate surveys. The sheer amount of data is staggering, and even individuals with extensive file management experience may not know where to begin.

When I received the 1979-1997 Surveys, the 1999 Surveys, the 2001 Surveys, and the 2003 data from the original principal investigators, I had several advantages altogether that I suspect several other researchers lacked. First, I had considerable experience with the substantive area of Public Understanding of Science. This was critical because it meant that I was already familiar with many of the questions on the Surveys, many of the disciplinary constructs, and other related research (e.g., Virginia Commonwealth’s Life Sciences surveys, see Funk, 2003 for an example). I knew about issues emanating from Sociology, Anthropology, Psychology, Political Science, Communication and Education.

All of this knowledge made it much easier for me to prioritize and select questions for “the core” of the surveys. For example, I knew which questions had been repeated (currently all items in the archive consist of questions repeated over at least two surveys). I had some notion of the amount of educational detail available and which questions could be compared with some of the other major surveys (e.g., the series on funding priorities with the General Social Survey). Thus, when creating a longitudinal database, I strongly recommend that the team include at least one individual with expertise in the data topic.

Second, because of my prior experience creating smaller longitudinal archives as well as constructing a quantitative data set from ethnographic material, I knew from the start how difficult it would be to construct an archive containing all the data from the NSF Surveys within a relatively short time period to work. I already had the experience of being overwhelmed by sheer amounts of data. For example, in the early 1990s I directed a field study of 38 North Florida religious congregations, with three months of observational data per church or synagogue (this produced over 14 linear feet of notes, documents, and interviews and a 20 single-spaced page codebook). Very early in the coding process, it became clear that completing the 20-page codebook for each congregation could take several years. My students and I switched to coding smaller units from the codebook (e.g., material on group cohesion or demographic composition); this made the entire coding process smoother and faster, producing a datafile with over 200 variables per congregation in less than a year.

Thus I knew that starting “small” in comprehensive units, and then joining these units together was a workable strategy. I had also created multiple year indices from other survey data, including the GSS. It probably seems obvious, but tackling a set of databases to combine them into a longitudinal archive is not a task for novices.

Third, prior to constructing the archives, I had taught introductory statistics for several semesters using California-Berkeley’s online Documentation and Statistical Analysis program (DAS; formerly “SDA” or “Survey and Documentation Analysis”). This speedy Internet statistics program (it tears through 135,000 cases in nine seconds—on dialup) encompasses most introductory and intermediate level statistical procedures. It is relatively straightforward and simple to use, with an extensive online users’ manual. Even introductory statistics students enjoy using it! From the beginning, I had a vision of creating an NSF Surveys archive in which not only could the data be placed online for downloading, but also an online experience in which scholars could conduct preliminary analyses online using the DAS system so they could decide in advance which variables or years they wanted to download using SPSS or SAS files. Knowing where you want to go helps provide a template for creating such an archive.

Fourth, I believe that online data archives form much of the future of analytic research and I have learned as much as possible about them. When I taught our department’s basic graduate methods course in the early 2000s, I devoted the last two weeks and the final student assignment to discovering online archives, questions to ask about their construction and use, and describing some of the wealth of materials already accessible on the Internet (see: ). Thus, I had several prior valuable experiences before creating the NSF Surveys archives: substantial disciplinary knowledge; constructing multi-year data files; creating quantitative files from extensive qualitative materials; working with online archives; and using the DAS statistical system.

Getting started: the IT file

Where does one begin with hundreds of survey variables, a dozen different surveys, different methodologies, and some questions asked in very different ways over a 25-year period? I took as my model the General Social Survey (Davis & Smith, 2004), with which I have extensive analytic experience. The eight characters or less mnemonics for the variable names tend to be intuitive (e.g., EDUC for years of education or DEGREE for highest degree level attained and I have incorporated these), each variable creates a column with standard inapplicable codes for years when a question is not asked and relatively standard sets of missing data codes, including the option to move to a two-digit system (e.g., 98 for “I don’t know”).

SPSS can be a slightly fussy program for creating archives (see below) but one of its traditional strengths is file management—creating new variables, collapsing variables, and recoding data. The archive creator who can write SPSS syntax can work more quickly still, especially in the case of variables such as detailed occupation, which can have hundreds of values. A complex syntax file can be created, then recalled and run sequentially on successive years of the data. On the other hand, when adding new years (i.e., joining files), SPSS demands identical codes (spelling, case, etc) and exact variable orders, which can make new entries tedious unless an earlier created syntax file is run.

Colleagues marvel that I use an ancient SPSS version to create these files: SPSS 6.1.3, the very last Windows 3.1 version, and it is true that there are times it can be positively “clunky”[1] and it lacks many of the file management capacities of more recent editions. On the other hand, the 6.1.3 on-screen menu to create new variables and enter parameters in them is much simpler than, for example, SPSS 14.0, and, most important, all later versions of SPSS are able to read the data, thereby increasing its utility internationally and for scholars who do not have the most recent versions of the program.

The first file I created addressed information technology, because, unlike some of the highly subjective and open-ended questions, the basics of IT are relatively concrete, thus making this topic an easier starting point. One has a computer or does not; an individual has an email account or does not. The number of hours one spends online can be estimated. Further, there were a limited number of questions in the NSF Surveys about IT (total under three dozen) compared with hundreds of diverse dimensions of science items. The IT items and basic educational and demographic variables initiated the first archive, and ultimately formed the nucleus of what became known as the large “megafile” or “the big one” of public understanding of science.

Because I became intrigued by gender, educational and occupational digital divides (Losh, 2004), I wanted the information technology series to continue. The IT series in the NSF Surveys ended in 1999; however, I knew that the 2002 GSS had modules on computer and Internet access and use.[2] Thus I “grafted” most of the comparable IT items from the subsample of the GSS that either reported having a telephone or who “didn’t know” (an unlikely occurrence), which comprised about 95 percent of the total[3]. I am now examining a more limited number of items in the 2004 GSS about IT access and use. Upon my initial inquiry, the University of Maryland’s WebUse site welcomed the “hybrid” survey, converted it to an earlier version of the DAS statistical program (see and scroll down to the “Miscellaneous” section), and was the first archival collection to place the data online.

The “Big One”

It took two more years to prepare the much larger “megafile”. I began calling the SPSS core Public Understanding of Science datafiles “KNOWX.sav” (short for knowledge) as I built and added to them. KNOW20 became the archive for the ICPSR ( see Study 4029) and the Roper Center ( enter “science” at the search engine) archives. I followed the same general procedure as I did constructing the IT file, creating smaller units, then joining the years together, then joining the multi-year units together. Some of the units include: basic demographics;[4] detailed educational information (e.g., exposure to high school biology, physics, chemistry, advanced math; college major, type of college degree, e.g., vocational versus liberal arts), basic science knowledge and inquiry understanding; pseudoscience support; IT access and use; self-rated interest and informedness on several topics, including science and technology; attitudes toward science, including stereotypes about scientists; funding priorities; political involvement; cultural engagement, including facilities (libraries, museums); and detailed media use.

Methodological considerations

Creating a massive archive involves more than simply creating clusters of variables and joining them together in SPSS files. The data present several challenges because they were collected over a long time period and, although the director was the same for 20 years, several of the methods changed. For example, the 1979 survey consisted of in-person interviews conducted by Temple University’s Institute for Survey Research. Six surveys (1981-1992) were Random Digit Dial interviews conducted by the Public Opinion Research Center at Northern Illinois University; four different agencies conducted RDD interviews between 1995 and 2001; the 2003 RDD telephone survey was mostly conducted by graduate students in the Universities of Maryland and Michigan joint program in survey methodology (JPSM), and NORC conducted in-person interviews in 2006 using probability area sampling. The times of the year differed (although the effect on these topical variables would not be expected to be substantial).

More immediate are discontinuities and differences in how questions were asked over the now 28-year period of science indicators. For example, detailed occupational classification was coded six different ways between 1979 and 2001, including an alphabetical listing (Actor, Baker, Carpenter, etc), and was not obtained at all in the JPSM 2003 survey. The most common codes used were the 1970 three digit detailed Census categories. Recognizing that I could not do a total match across the years, I created sets of dummy variables with a few variations on the Census collapsed occupations: science/engineering professional or manager; other professional; manager; technical (non-professional); clerical worker; or “other” occupation[5]; then ultimately I used the series of dummy variables to create a new variable called “OCCUP”. Crude? Yes, but it makes repeated cross-sectional analysis using an occupational variable possible that could not be done otherwise.

Variation in some media variables presented more challenges. In about half the years, respondents were asked:

“Now, I’d like to read you a short list of television shows and ask you to tell me whether you watch each show regularly, that is, most of the time, occasionally, or not at all.”

This is a recognition cognitive task, i.e., respondents either do or do not recognize the television program read and answer accordingly. Recognition is easier than recall, which is shown in the other series of questions about television viewing:

“Do you watch any television shows that focus primarily on science or nature? [IF YES:] Which science or nature show do you watch most often?”

These two versions of the questions, therefore, not only take different formats, but present respondents with two quite different cognitive tasks. These questions could not be collapsed so I created two separate “variable “streams”. In the codebook I created for the online archives, I urge analysts to:

“…cross-tabulate television program questions by year to ascertain which form was asked when before engaging in analysis.”

The situation for amount of time spent watching daily television news was even more complex: how much the respondent watched television news was asked several ways. In one (collapsed) set of questions, respondents were asked how many hours per day they watched the news. In the other major stream, respondents were separately asked whether they watched morning, evening, and late night news “regularly, occasionally, or not at all.” In addition to leaving the questions separate, I decided to see if there were parallels between these different sets of questions. If these existed, across-time comparisons could then be created.

Thus, I constructed an entirely new variable in which respondents who reported regularly watching all newscasts were coded as at least one hour daily; those who said they watched none of these newscasts were coded as none. Counting early evening news as one hour and the other two broadcasts as approximately one-half hour each, I coded different combinations as either less than one-hour daily or as more than one hour daily. The resulting new variable “NEWSHR3” was cross tabulated against literal time spent watching news for the 6094 respondents in four separate surveys who had been asked both sets of questions. The crosstabulation results indicated that the collapsed “NEWSHR3” created from the “regularly, occasionally, none” series approximated hours of watching for 92% of these 6094 adults. This correspondence meant that a new timeline on watching television news was now available to extend the number of years for analysis.

Other discrepancies in question wording or codes paled beside the occupational and media variables. In most cases, the differences amounted to minor spelling variations. One of the biggest annoyances was moving variables across the columns so that they would be compatible over the survey years.

Creating cohorts presents other problems. True “net” change, of course, cannot be assessed in synthetic cohorts since each survey brings a fresh crop of adult respondents. More serious to consider, however, in watching either natural or synthetic cohorts age is selective attrition. Individuals with more education live longer (in part due to better health information and practices), but they also generally have had more exposure to science courses (Losh, 2006a) and select more scientifically accurate media (Miller, 2007). Thus, it might appear that older individuals are more knowledgeable or reject pseudoscience more (see Losh, 2007) than younger ones—but this could be due to selective attrition rather than to any component of the aging process.

Going Online

If it isn’t clear by now, I think the NSF Surveys archive is a national treasure for scholars for years to come. But much of its future value lies in continuing the series. I reasoned that the more people using the data the more likely the series is to continue. Further, it is clear that the future of secondary analysis using data archives is online, where data can be readily accessible, downloaded, or even analyzed online for preliminary research findings. Michigan’s ICPSR received six million dollars in 2004 from the Library of Congress to increase its online archives.

Keeping user frequency and easy accessibility in mind meant looking for the most “worthy” online collections, preferably with DAS capacity. In early 2004, this narrowed the agenda to two: Michigan’s Inter-University Consortium for Political and Social Research, and the University of Connecticut’s Roper Center for Public Opinion.[6] Both have sizable archives, including many topics (not just public opinion or social research)[7] and both use the DAS statistical program (the Roper Center through its iPoll system).

Each archival collection has a Board of Directors, which reviews and approves additions to its collection. ICPSR has a questionnaire of several dozen pages[8] that gathers information about potential data sources for the archive (e.g., type of data collection, type of participants, response rates), principal investigators, use in conferences and publications, and other information. The Roper Center has a series of questions about the data (e.g., year, time of year, data gathering agency), which, ironically, did not overlap very much with the ICPSR questions.

Due to its entire conversion to a DAS file (in addition to the original SPSS file and SAS conversions), it was a year (late January 2005) before the NSF Surveys ICPSR datafile made its debut. ICPSR employed a new version of the DAS statistical system, which is a pleasure to use. Like a proud parent, I announced the archives’ arrival on several of my member list_servs. That’s when I found out that only faculty, staff and students at ICPSR “member institutions” can access both the IT and the “megafile” archives, download the data and use current versions of the DAS program to analyze the archives—and they can only do so from “approved” Internet Provider addresses. Because at the time the Florida State University Libraries had not sent ICPSR a list of these authorized addresses, for over six months the only way I could access the data was through FSU’s dialup system (FSU is an ICPSR member). Faculty or students off campus or traveling similarly could not access the data until our libraries’ distance ISP proxy was approved in Ann Arbor. The National Science Foundation could not (and still cannot) access its own data at ICPSR. These diverse restrictions are an instructive tale about questions one should initially ask about contributing a dataset to an online archival system.

When I contacted the Roper Center, I mentioned the issue of free access up front. The Roper Center Board of directors approved both the archive and free distribution, although users must register. Limited analysis can also be done using an earlier DAS system through its “iPoll” feature. Both Roper and ICPSR have extensive user-friendly documentation on their archives, and both upgrade the online DAS system as improvements become available. ICPSR has extensive bibliographic information about papers and publications produced from the data although Roper has many links to original sources and publications from them.

USING THE ARCHIVES

I want to give some idea here of how the data now are being used. Some publication information is online with the ICPSR archival site, which provides some bibliographic citations. I know the most about archival use by either academic scholars or government personnel. Jon Miller (e.g., Miller 1983; 1998; 2004; Miller & Kimmel, 1998), with his now-professional life expertise in CLS accounts for the overwhelming majority of bibliographic entries, although I am contributing my share.

Last August (2006), I spoke first hand with both ICPSR and Roper Center representatives. ICPSR pulled up several screens of well over 100 users who had accessed the files to date. The Roper Center reported there have been several hundred accesses, and apparently many faculty download the archives to use for course assignments in both science related (history and philosophy of science; public administration) and statistics courses.

There is considerable international interest in the data. At this point, topics such as genetically engineered foods have been more controversial in Europe than in the United States. On the other hand, through political rancor, Americans may be more involved with controversies over stem cell research, global warming, and the artificial prolongation of life than Europeans. U.S. citizens see biological evolution as definitely more controversial than most of the rest of the world, although this may be changing.[9]

The Eurobarometer surveys (e.g., Allum, et al., in press), which include sizable components on adult civic science literacy, now span well over a decade. There are several additional British surveys on CSL (e.g., Office of Science and Technology & the Wellcome Trust, 2000). Over the past few years, South Korea and India have begun science literacy indicator surveys of the adult general population and development is now underway in Brazil for a comparable series. The recent Public Communication of Science and Technology (PCST-9) conference in South Korea spotlighted the extensive interest in adult science and technology indicator surveys in Vietnam, Indonesia, and Africa. Scholars such as Martin Bauer or Nick Allum in Great Britain are already using the NSF Surveys archive in comparative analyses.

These international comparisons have become increasingly important in the science and technology knowledge “horse races.” For example, Miller (2007) not only presented data showing increases in adult American CSL over the past few decades but also reported that U.S. adults were second only to Sweden in basic CSL by the early 2000s. On the other hand, American students fare much worse in international lists. Miller (2004, 2007) attributes this student-adult differential in international science standing to the American college and university requirement of at least one science course for graduation. However, Losh (2006), who also finds an increase in U.S. adult CSL over the generations, points out that most Americans do not even have a two-year college degree—but that more recent birth cohorts do show huge increases in high school biology, chemistry and physical science courses. There are also, of course, large international differences in which students take the high school achievement tests that form the basis for cross-national comparisons. However, nearly all the adult studies use comparable general population samples, which many would argue are more pertinent for comparative international research. In any event, the stage is set for the continued development of large-scale national databases and further international comparisons to probe civic science and technology literacy. Thus archives such as the ones I created are certain to proliferate in the future.

REFERENCES

Allum, N., Sturgis, P., Tabourazi, D., & Brunton-Smith, I. (2007, In Press). Science knowledge and attitudes across cultures: a meta-analysis. Public Understanding of Science.

Bauer, M., J. Durant & G. Evans (1994). European perceptions of science. International Journal of Public Opinion Research, 6 (2:) 163-186.

Bauer, M.W., Allum, N., & Miller, S. (2007). What can we learn from 25 years of PUS survey research? Liberating and expanding the agenda. Public Understanding of Science, 16: 79-95.

Burris, J. (2006) Testimony offered to the Research Subcommittee of the Committee on Science of the U.S. House of Representatives Hearing on Undergraduate Science, Math & Engineering Education: What’s Working? March 15.

Davis, J.A., & Smith, T.W. (2004). General Social Surveys, 1972-2004: Cumulative Codebook. National Opinion Research Center. Principal Investigator, James A. Davis and Co-Principal Investigator, Tom W. Smith; Co-Principal Investigator, Peter V. Marsden. NORC ed. Chicago: National Opinion Research Center, producer. The Roper Center for Public Opinion Research, University of Connecticut, distributor.

Funk, C. (2003). VCU life sciences survey: Public values science but concerned about biotechnology.. Retrieved from source.

Lemonick, M., Keegan, R.W., & I. Ybarra (2006). Is America flunking science? Time, February 13, 23-33.

Losh, S.C. (2003). Gender and educational digital gaps: 1983-2000. IT & Society, 1 (No. 5, Summer:) 56-71. . Retrieved from source.

Losh, S.C. (2004a). Gender, educational and occupational digital gaps: 1983-2002, Social Science Computer Review, 22 (Summer:) 152-166.

Losh, S.C. (2004b). Science. Pp. 317-326 in J. Geer (ed.), Public Opinion and Polling around the World: A Historical Encyclopedia. Santa Barbara CA: ABC-CLIO, Inc.

Losh, S.C. (2006a). Generational and educational effects on basic U.S. adult civic science literacy. Pp. 836-845 In The Korean Science Foundation and the Korean Academy of Science and Technology: Proceedings of The Ninth International Conference on Public Communication of Science and Technology (PCST-9).

Losh, S.C. (2006b). American Stereotypes about Scientists: Gender and Time Effects, Annual Meetings of the American Sociological Association, Montreal, August.

Losh, S.C. (2007). Age, generation, and American adult support for pseudoscience. Paper presented at the annual meetings of the American Association for the Advancement of Science, San Francisco, February 17.

Losh, S.C., Tavani, C.M., Njoroge, R., Wilke, R., & M. McAuley. (2003). What does education really do? Educational dimensions and pseudoscience support in the American general public, 1979-2001. The Skeptical Inquirer, 27, 5: 30-35.

Miller, J.D. (1983) Scientific literacy: A conceptual and empirical review. Daedalus, 112 (2:) 29-48.

Miller, J.D. (1998) The measurement of civic science literacy. Public Understanding of Science, 7 (3:) 202-223.

Miller, J.D. (2000). The development of civic scientific literacy in the United States. In Kumar, D.D. Chubin & D.E. Chubin (eds.) Science, Technology, and Society: A Sourcebook on Research and Practice. New York: Kluwer Academic/Plenum Publishers: 21-47.

Miller, J.D. (2004). Scientific literacy in the United States: The linkage between schooling and adult skills. Paper presented at the Annual Meetings of the American Educational Research Association, San Diego, April 12.

Miller, J.D. (2007). Civic science literacy across the life cycle. Paper presented at the annual meetings of the American Association for the Advancement of Science, San Francisco, February 17.

Miller, J.D. & L. Kimmel (1998). Science and technology: public attitudes and public understanding. Chapter 7 (1-22) in National Science Board, Science & Engineering Indicators—1998. Arlington, VA: National Science Foundation (NSB98-1).

Office of Science and Technology & the Wellcome Trust (2000). Science and the Public: A Review of Science Communication and Public Attitudes to Science in Britain. United Kingdom: Office of Science and Technology.

Seymour, E. (2006). Testimony offered to the Research Subcommittee of the Committee on Science of the U.S. House of Representatives Hearing on Undergraduate Science, Math & Engineering Education: What’s Working? March 15.

U.S. Bureau of the Census (2006). The Statistical Abstract of the United States. Washington, D.C.: U.S. Government Printing Office.

Weiman, C. (2006) Testimony offered to the Research Subcommittee of the Committee on Science of the U.S. House of Representatives Hearing on Undergraduate Science, Math & Engineering Education: What’s Working? March 15.

-----------------------

[1] For example, to create a new folder in 6.1.3, one must enter Windows Explorer (or other file management window), create the folder there, and then return to SPSS to save a file within the newly created folder.

[2] Many IT questions were also asked in the 2000 GSS but problems with branching and skip patterns prevented making general population inferences from the data.

[3] A similar subsample with telephones will be extracted for the 2006 GSS Public Understanding of Science module.

[4] Race/ethnicity is available from 1999 forward; USA nine-census region information is available from 2001 forward.

[5] More detail on blue-collar occupations is yet to come.

[6] I would now add the Pew Research Center, which has recently increased the variety of its holdings.

[7] For example, several years of NAEP data are in the ICPSR holdings.

[8] My quip has been that it took longer to complete the questionnaire than to create the megafile archive!

[9] “Biblical creation” as a challenge to Darwinian evolution is becoming more common in both England and France, and Turkey may drop evolution as an approved and required topic in public school science classes.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download