Abstract

Government statistical agencies such as the Census Bureau go to great lengths to make sure data collected are kept strictly confidential, and sharing of data between government agencies is severely restricted. Discussions of confidentiality of business data typically ignore concerns about the burden placed on companies that supply data to multiple government agencies. From the perspective of many large companies, however, burden is of great concern since these companies, because of their size, are respondents in multiple surveys. This paper describes findings from qualitative research on the topic of confidentiality and reporting burden from the perspective of 30 large companies. Companies acknowledged their data are confidential, but much of what we ask typically has been released earlier to shareholders and upper management. Companies put much more emphasis on the number of survey requests placed on them, since many of them seemed to ask for 'similar information' from their perspective. Large companies generally supported data sharing among statistical agencies under well-specified conditions and with rigorous security and confidentiality provisions in place. They only saw value in this sharing, however, if it reduced the reporting burden placed on them.

Key Words: Confidentiality, Respondent Burden, Business Surveys, Qualitative Research

Introduction

Recent media coverage has highlighted instances where information about individuals has been shared by private sector organizations with other organizations without the individual's consent. The resulting growing concern among the general public about the sharing of personal information could possibly affect participation in government agency statistical data collections, even though there are no documented instances of illegal data sharing by these agencies. Research has shown that promises of confidentiality improve cooperation, and methodologists have argued that breaches of confidentiality would lead to significantly reduced respondent cooperation (Dillman et al., 1996; Meyers & Oliver, 1978; National Research Council, 1979; Reamer, 1979; Singer, 1978, 1993; Singer, Von Thurn & Miller, 1995).

At the U.S. Census Bureau most data are collected under strict confidentiality agreement, with legal protections under Title 13, section 9. This prohibits dissemination of data in a manner that allows identification of the respondent or is in any way be harmful to him as defined by the President's Commission on Federal Statistics (1971). This prohibition pertains to nonsworn individuals in the public and in other government agencies. Other government and nongovernment data collection agencies use variations of similar agreements (Federal Committee on Statistical Methodology, 1994). Typically, confidentiality laws that govern large governmental statistical agencies do not distinguish between individuals, households and establishments. As an example, U.S.C. Title 13 which governs most Census Bureau surveys makes no such distinction.

Government data collection agencies including the Census Bureau and the Bureau of Labor Statistics (BLS) devote significant resources to maintaining confidentiality of the data collected. Much of the literature has focused on methods for keeping business and household data confidential, such as the use of restrictions, disclosure limitations and laws (Bulmer, 1979; Butz, 1985; Choldin, 1988; Cox et al., 1985; National Research Council, 1993; Cecil, 1993; Federal Committee on Statistical Methodology, 1994).

Unfortunately, studies conducted by the Census Bureau confirm that the public either is unaware of the agency's stringent confidentiality standards or do not believe them (Clark, internal memorandum, 1997; Singer et al., 1997; Gates & Bolton, 1998). Much of this research is limited to individual and household level data collection. As a byproduct of two economic response behavior studies, we've gathered some information on knowledge of census confidentiality rules from small business firms. Approximately half of the small single-location firms surveyed in December 1987 prior to the 1987 Economic Census thought that census data about individual companies were kept confidential and 30% claimed not to know whether they were kept confidential or not. Even after promotional and outreach activities attempting to increase knowledge, knowledge of confidentiality rules increased only modestly, to around 60%, with 20% still unsure whether data are kept confidential (Zeisset et al., 1990). Similar percentages were found by Vacca (1990) who surveyed farmers about the 1987 Census of Agriculture.

Ethicists claim, and the courts have agreed, that individuals have a right to privacy. Moreover, there is clearly some information that even large businesses consider proprietary, such as the development of new products and services or strategic plans. Efforts by governmental regulatory agencies or competitors to obtain such information have clearly been thwarted by the use of provisions in Title 13 (Corcoran, 1963). Such information, however, is not the kind of information normally obtained by the Census Bureau or other government statistical data collectors. Certainly, other competitive information collected by the Census Bureau, such as revenue from individual stores, is also considered confidential by the company. The current confidentiality rules the Census Bureau maintains, which apply to all topics and all reporting units, whether they are businesses or individuals, is the most conservative course. A blind application to businesses of the same confidentiality rules and procedures that the Census Bureau currently applies to individuals fails to pay sufficient attention to possible differences between the two - differences which might argue for different confidentiality treatments.

One important difference is the level of public information available about businesses, as compared to individuals. Publicly owned companies are required to publish substantial information about their business. This information is provided, among other ways, in the 10K reports that the business submits to the Securities and Exchange Commission (SEC) and is available publicly. Thus, some data on businesses collected for regulatory purposes are considered public and are not collected under a pledge of confidentiality (Federal Committee on Statistical Methodology, 1994). Very often this level of information is available on company Web sites, making the availability worldwide and relatively simple to access. Even private and small firms have Web sites, but these do not typically reveal the level of detail provided by public companies. Edwards and Cantor (1991) suggest these latter businesses (individual proprietors such as farm operators) are more like individuals and households in the level of data maintained out of the public eye, and perhaps have some of the same privacy concerns as individuals.

Another distinction between businesses and households is their required level of participation in survey requests. An unintended consequence of the confidentiality regulations is that a business may be required to provide the same, or virtually the same information to multiple government agencies. For example the current application of the law prevents building a common business frame across statistical agencies (Jabine, 1993), and requires several agencies to independently obtain listings of establishments. The burden is especially onerous for large businesses because, for reasons of sample efficiency, they are included with certainty or with very high probabilities in every applicable survey across statistical agencies. This fact was recognized by the National Performance Review (1993), which urged the elimination of barriers to the exchange of business data among federal statistical agencies, especially the Census Bureau, the BLS and the Bureau of Economic Analysis (BEA). Indeed, some limited exchange of information between the Census Bureau and other agencies is already permitted (Trott, 1995; Jabine, 1993).

In contrast, for respondent burden reasons, the Census Bureau makes a great effort not to select a household for more than one demographic survey. Of course all households are in the decennial census, and many need to complete a tax return annually, but generally a household's probability of selection into a government statistical agency survey is relatively low. Requested participation in more than one survey is almost unheard of. Thus, imposition of greater reporting burden which affects businesses, especially large businesses, and is due in part to current confidentiality regulations, does not affect households to the same extent.

Aside from the ethical issues of data confidentiality for large companies, there is the practical question of whether promises of confidentiality increase cooperation. For instance, rather than increasing cooperation, maintaining our current level of confidentiality (which restricts data sharing) may decrease cooperation from large businesses by increasing reporting burden. As indicated in the previous paragraph, respondent burden is less highly related to confidentiality for individuals, although even for them it is noted that cooperation is reduced if the introductory statements, which typically contain confidentiality assurances, are time consuming and thus burdensome (Groves & Couper, 1998).

Business Attitudes Toward Confidentiality & Burden

As a first step in research to understand and reduce respondent burden in large multi-unit businesses, the Census Bureau conducted unstructured qualitative interviews with corporate office data providers from 30 large multi-unit businesses during 1998-99. The initial study was limited to large businesses since they account for a very high fraction of the estimates reported by the Census Bureau and are believed to face more complex statistical reporting issues than smaller companies. The 30 businesses were selected judgmentally to represent a broad range of industries and other characteristics of large businesses. Most were publicly owned and report to multiple governmental taxing and regulatory agencies, including the Internal Revenue Service (IRS) and SEC. A range of issues was discussed, including the businesses' organizational structure, the availability of requested data in their information systems and more detailed questions pertaining to the reporting process and specific Census Bureau survey requests. These were all framed within the context of the visit, to find ways to reduce reporting burden. We also discussed the companies' perception of our confidentiality rules, the confidentiality of the data we request, and how this relates to the different requests made on them.

We caution that the results of this research are qualitative. More research is required to confirm the results we report. In addition, these companies do not represent medium and smaller-sized businesses, which may have different concerns such as the relevancy of the data collection to themselves and their business. Also, the main focus of the meeting was not on confidentiality; although we mentioned the topic in the agenda we sent prior to our visit, the time devoted to that discussion was limited. Given these disclaimers, the following responses were consistent across companies visited and demonstrate relationships between maintaining confidentiality and possible burden reduction.

Repeatedly, data providers in these large businesses complained about duplication of effort and the burden of having to provide the same information to multiple government agencies. They made comments like, "How many times do we have to report our net revenues?" The perception of duplicate data requests was common, especially among public companies with many outside data requests placed on them. Typically, the responsibility for providing the data falls on a unit within the Accounting, Finance or Controller's Office. This unit responds to internal management requests, external requests placed by statistical agencies, and, if applicable, the SEC, the Federal Trade Commission (FTC), stock analyst and trade association requests.

Each survey is seen as adding to the burden. The common question asked was, "Why can't you guys get your act together?" The "you guys" comment was directed toward "the government" in general and not just the Census Bureau. Companies did not appear to distinguish between the Census Bureau, Energy Information Administration, BLS, BEA, and the Department of Transportation. To manage this activity, a few companies compile a list of surveys and survey sponsors. Very often a survey from one agency was incorrectly attributed to a different agency. This held especially true for agencies under the same department, such as BEA and the Census Bureau, which are both within the Department of Commerce. More than one company was surprised when told that BEA and the Census Bureau did not regularly exchange data. Companies distinguished between government agencies like the Census Bureau that gather information for statistical purposes, and agencies like IRS that gather data for taxation purposes, most likely because another group of company accountants completed IRS requests, and because there are different penalties associated with these requests. Another regulatory agency, the SEC, was also seen as a different entity, even though the company staff with whom we met were the people who completed that request.

This "government" view also became apparent when we asked the question, "Do you believe that all government agencies use the same confidentiality rules?" It was obvious that none of these companies had ever thought about possible differences between the confidentiality rules. Often, company respondents offered a guess and we had to assure them this was not a trick question. Companies thought we used the same confidentiality rules or they claimed not to know, both of which implied that these companies did not have a full understanding of the Census Bureau's or other agencies' confidentiality pledges. One company hoped the rules were the same.

This is not to say that companies did not claim to have a basic level of trust in the Census Bureau's confidentiality pledge. When asked, "Do you have any concerns about the level of confidentiality the Census Bureau currently maintains?" most often the answer was no. Frequently, however, companies offered their response along with the comment that they hadn't heard of any data leaks. It was obvious that many respondents had not thought about our confidentiality rules before our visit, and in one company, the printed rule had to be pointed out to the respondent. In addition, some companies acknowledged (especially for our monthly or quarterly indicator reports) that they actually hold up release of the data to the Census Bureau until the release of that data to management and/or stockholders (if applicable). Even though they claim not to have concerns about the level of confidentiality maintained, actions like holding up a data release indicate some level of discomfort with or ignorance of our confidentiality rules.⁽²⁾ At the very least these companies were very prudent and would much rather be safe than sorry. This policy of holding up release of data, however, did not pertain just to government data requests. Rather, it was enacted across all data requests. The companies' first priority was to make sure audited figures agreed and that management had time to review them. Once data had been reviewed and distributed to the priority customers, like management and stockholders (for public companies), the data were considered public to an extent and could be released to others.

Most companies acknowledged that the type of information we request is confidential, but much of it is typically something they provide to others, either management and analyst requests or other regulatory requests. A typical comment was, "Most of the data the Census Bureau wants we already provide to the SEC and in our annual reports." So when asked, "Are there any data that you do not provide [to the Census Bureau] because your company considers this information to be confidential or proprietary?" the typical response was that if the Census Bureau requests the data, and the company has the data, they report it, especially if it is already in the public domain. Sometimes the detailed data required in some censuses and surveys could reveal information about a company's strategic planning. While this level of information is available to management, it is not usually available publicly. For instance, retail companies consider individual store revenue of strategic importance. Contracting/consulting companies consider rate structures for services or profit margins confidential. Although the Census Bureau does not specifically ask for these type of data, similar data (e.g., gross margins) could possibly be calculated indirectly or modeled by some of the information collected.

Requests for very detailed information are problematic for other reasons as well. The very detailed data are often unavailable from the businesses' corporate information systems. As a rough rule of thumb, summarized information based on data available from the business' general ledger, used to create income statements and balance sheets, is both easy to obtain and considered fairly public. Information from sub-ledgers or individual locations is more difficult to obtain and is more likely to be considered confidential. Only a few companies claimed they do not have a problem with our confidentiality rules because the Census Bureau does not request information of a strategic importance.

Greater concern about confidentiality was expressed by companies in very competitive markets and from two companies that had past internal data leaking experiences (not from the federal statistical agencies). Interestingly, the relatively few (i.e., four) large private companies visited provided the same type of responses presented by the public companies. Although one was extremely concerned about revealing information of strategic importance and was protective of what was said, even during the visit, others adopted an attitude similar to the public companies. Even private companies are subjected to some federal regulations, requiring reporting to regulators and to debt holders. Such data are not considered confidential. It is difficult to generalize for private companies, however, since we visited only four.

Even when there were some concerns expressed about confidentiality, companies generally placed much more stress on reporting burden issues. As our final question during the confidentiality portion of our visit we asked, "To reduce reporting burden on your company, it has been suggested that multiple government agencies share common data collected from a company, which means you would provide the data only once. What are your thoughts about that? What protections would you want to see if such a system were established?" Large companies generally supported data sharing among statistical agencies under well-specified conditions and with rigorous security and confidentiality provisions in place. To them, it is a single federal agency. They only saw value in this sharing, however, if it reduced the reporting burden placed on them. This sentiment influenced two of the four companies that did not support the data sharing idea. Neither one experienced a lot of duplicate data requests. Smaller firms might not see as large a benefit to data sharing for the same reason.

Companies acknowledged that the wider the access, the greater chance there is of leaks and of misinterpretation. To make data sharing work, companies requested several restrictions. The data sharing should involve strict security pledges. The requests need to be coordinated so the different agencies ask for, use, and interpret the data in the same fashion. The company comment, "You got five people that want it five different ways" highlights the problem of different agencies asking for the same general data, but in slightly different ways. It is clear that there is a company perception of duplicate data requests, perhaps even when agencies correctly claim there is not a lot of exact duplication.

Companies were cognizant of the necessity for a government infrastructure to support such an endeavor. This would include coordination of systems to access and use data, consistent formatting of data for ease of compliance, and detailed metadata to ensure data are used correctly and eliminate questions concerning the data provided.

Companies wanted to be kept abreast of any changes to the agreement (e.g., new agencies gaining access to the data), to restrict the agencies with access to the data (e.g., no IRS), and the purposes for which the data would be used. They were perfectly willing to have the statistical agencies share their data, but did not want these data shared with taxing agencies, and expressed mixed feelings about sharing with other regulatory agencies like the Federal Trade Commission, since different staff often provide information for statistical and legal purposes.

Even though in theory, companies thought statistical agency data sharing was a great idea, they acknowledged it will take work to implement. Several mentioned that different units within their own organizations request similar data from their group. They were internally plagued by the same duplicate data request problem that the federal government imposed on them.

Ideas for Discussion

At the end of one of our meetings, one company summarized the current situation, "... you mentioned confidentiality and the sharing, what do we need to do to get that moving?" This section presents some of the authors' ideas. We are no more than lay observers, who cannot speak officially for the Census Bureau as to whether these ideas could, would, or even should be pursued because of implementation and legality issues. In addition, these ideas did not originate from the companies, although in some cases the implementation issues associated with them did.

First, we point out that steps can be taken to reduce reporting burden which do not involve issues of confidentiality or sharing across agency. If data are accessible on the Web, theoretically, a company could grant several organizations access to the same password- protected Web accessible data, thereby reducing reporting burden. We are well aware of the implementation challenges raised by this approach. Reporting burden is not necessarily reduced, if the data set is accessible, but the company had to create it with same number of items normally requested separately. Preliminary work would need to be done to establish common definitions, reference periods and variables so that the number of data items requested would actually be reduced. We also speculate there might be confusion if the metadata associated with the data set was not thorough. Organizations could end up using the wrong item and/or calling the business to verify the data. These result in either poor estimates and/or more burden on the business. Most likely, one statistical agency would be in charge of coordinating this request. Which one? Finally, the different agencies would need to have appropriate hardware and software to access the data.

As was apparent during preparation for our visits, the Census Bureau often makes multiple data requests to large companies. It makes sense for each agency to "inventory" their own surveys for any possible duplication of data items (such as payroll and employment) across requests and then to conduct an inventory across agency. Investigating further the business respondents' perception of duplication would identify the most problematic variables. Recently the interagency Cognitive Applications for Establishment Surveys Working Group (2001) studied how the same concept was operationalized in survey questions and instructions in four surveys at BLS and three surveys (or censuses) at the U.S. Census Bureau (Fisher et al. 2001). The variables studied were total employment, number of production (nonsupervisory) workers, and production (nonsupervisory) worker hours. While the core definitions for the employment concepts were fairly consistent, the two agencies differed markedly in how these definitions were conveyed to respondents in survey questions and instructions, and different strategies for collecting these data were used. In addition, "hours" were defined as either "hours worked" or "hours paid," depending on the survey and the collecting agency. As expected, the reference periods for the surveys varied - one survey was monthly, the others were annual - and the reference dates were typically framed in terms of the pay period which included the 12^th of a particular month. These variations in question wording, order and structure certainly would contribute to business perceptions of "You got five people who want it five different ways."

In an attempt to evaluate whether the differences in question-asking strategies impacted the data collected, manufacturing sector data series for these variables were compared across the two agencies. Although the various series required some manipulation in order to compare them, minimal differences were found in estimates of total employment and the hours worked/paid ratio (for most industries, they were within �5%). However, differences in estimates of the number of production workers were much more variable, and the reasons were indeterminate. This study shows that the underlying definitions of several basic economic concepts are the same across surveys/agencies. Evaluating more data items from more agencies would be useful. Measurement error studies in which individual survey responses are compared across surveys and agencies for a matched sample of businesses would measure the impact of different questioning strategies on reported data. This would help identify where modifications could be made in questions/instructions to reduce respondent burden, while having minimal effects on estimates.

Another possible avenue to reducing reporting burden is through the use of administrative records. For small companies, the Census Bureau already obtains data from administrative records. Based on our visits, the business' perception is that much of what they provide is already public, either on the Web page or provided to the SEC. Here too there is some implementation difficulty. Although some of what the Census Bureau requests could perhaps be found publicly, there certainly would be metadata issues associated with that information, especially distinguishing foreign from domestic components for these large companies. Secondly, public data sources do not contain the detailed information that is typically found in some Census Bureau requests for multi-unit businesses, such as sales, employment or payroll for individual locations/establishments. We suspect that eliminating these data would have implications for important economic indicators.

Another idea is to implement more traditional notions of data sharing across statistical agencies. Allowing data sharing of all or selected large public business economic data among federal statistical agencies does, in theory, appear to potentially benefit the data providers by reducing their burden. As our visit discussions indicated, these large companies see the potential benefits of data sharing in terms of reduced reporting burden and appear to be willing to work with the federal agencies to achieve it. As indicated earlier, we speculate there is not as much advantage in this for individuals and households, small and medium-sized businesses, who might therefore not be as willing participants. In fact, knowledge of data sharing might potentially have harmful effects on these respondents in terms of reduced cooperation rates. If large business data sharing were to take place, studies are needed to determine whether this would benefit the data collectors by increased cooperation, timeliness, and accuracy. Studies are also needed to insure changes in economic data sharing policy do not have a negative impact on household cooperation due to media misperceptions of our policy.

The most obvious implementation issue is how to achieve data sharing legally. BLS and the Census Bureau abide under different confidentiality provisions, where the Census Bureau cannot share data without changes in the law. The BLS however operates under a Commissioner's Confidentiality Order, which has been cited in withholding information. Its intent is to protect the confidentiality of information received from the states. Both agencies also protect information under the Privacy Act. Nevertheless, as we noted, businesses believe that all government statistical agencies abide by the same confidentiality laws. It is our understanding that two years ago, a proposed legislative bill, H.R. 2885, the Statistical Efficiency Act, sought to enhance the confidential treatment of information provided to Federal statistical agencies and facilitate the sharing of information among those agencies for statistical purposes. Another active proposal would extend confidentiality protections to all federal statistical agencies, but not providing for data sharing. None of them has, as yet, become law. The likelihood of any approval of data sharing in the near future is uncertain.

There is a need to acquire industry champions to support the legislation. Additionally, the opinions and attitudes we obtained from 30 companies is most likely the type of information needed on a larger scale in order to pass any new legislation. In addition to knowing the types of data which are, and are not, of strategic importance or confidential, we need to know whether there are particular time periods for which businesses consider data to be no longer confidential or of strategic importance. This type of information could also help shape any changes to the confidentiality laws for future data collections.

Since we are not legal experts, we question whether there would be a need for changes in the confidentiality laws. Could data sharing of business data among federal agencies take place if such sharing is specifically requested by that business? This would most likely require a written consent or waiver by the company. Jabine (1993), however, cites two Census Bureau examples in which OMB refused to grant release of individual information collected under Title 13, even though a waiver was obtained.

If data sharing were to take place, the companies we visited indicated several implementation issues to be addressed in order for them to willingly take part. As with the password-protected Web site idea, there still would need to be coordination across agencies so that the same variables are requested, there would be a need for a government infrastructure to facilitate the data sharing, and there would be a need for good documentation or metadata. In addition, since the data sharing would no longer be in the companies' control, they requested (1) new laws or regulations which specify the level of security and confidentiality that applies to anyone with access to the data; (2) joint agreement concerning access to the data; (3) clear communication to the companies of these parameters and any future modifications; and finally (4) proof to the companies that this does indeed reduce reporting burden.

The status quo, which stresses confidentiality at the expense of respondent burden, is clearly not preferred by most large businesses who are basically responsible for the overall quality of economic data obtained. A solution that reduces burden needs to be found. At the very least, our work to date suggests that researchers cannot treat confidentiality as a single, undifferentiated concept, but will need to distinguish between individuals and businesses, between larger and smaller businesses, between public and private businesses, and by types and timing of data requested.

References

Bulmer, M. 1979. "The Impact of Privacy upon Social Research" in Bulmer, Martin (ed.) Censuses , Surveys and Privacy. New York: Holmes and Meier Publishers.

Butz, W. 1985. "Data Confidentiality and Public Perceptions: The Case of the European Censuses. Proceedings of the Section on Survey Research Methods, American Statistical Association

Choldin, H. 1988. "Government Statistics: The Conflict Between Research and Privacy," Demography 25, pp. 145-54.

Clark, C. and Miller, M. C. "Confidentiality and Privacy Research Status," Internal Memorandum for Martha Farnsworth Riche, U.S. Bureau of the Census, October 2, 1997.

Dillman, D., Singer, E., Clark, J., and Treat J. 1996. "Effects of Benefits Appeals, Mandatory Appeals and Variations in Statements of Confidentiality on Completion Rates for Census Questionnaires," Public Opinion Quarterly 60, pp. 376-89.

Edwards, W. S. and Cantor, D. 1991. "Toward a Response Model in Establishment Surveys," in Biemer, Paul et al. (editors), Measurement Error in Surveys. New York: John Wiley & Sons, pp. 211-233.

Federal Committee on Statistical Methodology. May 1994. "Report on Statistical Disclosure Limitation Methodology" Washington, D.C.: Office of Management and Budget, Statistical Policy Working Paper 22.

Federal Statistics: Report of the President's Commission on Federal Statistics, Vol. 1. 1971. President's Commission on Federal Statistics. Washington, D.C.: U.S. Government Printing Office, p. 222.

Gates, G. and Bolton, D. 1998. "Privacy Research Involving the Expanded Uses of Administrative Records", 1998 Proceedings of the Government and Social Statistics Sections, American Statistical Association.

Groves, R. and Couper, M. 1998. Nonresponse in Household Interview Surveys, New York: John Wiley & Sons, Inc.

Jabine, T.B. 1993. "Procedures for Restricted Data Access," Journal of Official Statistics, Vol. 9, No. 2, pp. 537-89.

Meyers, S.M. and Oliver, J.D. 1978. "Privacy and Hostility Toward Government as Reasons for Nonresponse in the National Medical Care Expenditure Survey," Proceedings of the Social Statistics Section, American Statistical Association.

National Performance Review, 1993. Creating a Government that Works Better and Costs Less: The Report of the National Performance Review. New York: Penguin Books.

National Research Council, 1979. Privacy and Confidentiality as Factors in Survey Response. Washington D.C.: National Academy Press.

National Research Council, 1993. Private Lives and Public Policies. Washington D.C.: National Academy Press.

Reamer, F.G. 1979. "Protecting Research Subjects and Unintended Consequences: The Effect of Guarantees of Confidentiality, Public Opinion Quarterly 43, pp. 497-506.

Singer, E. 1978. "Informed Consent: Consequences for Response Rate and Response Quality in Social Surveys," American Sociological Revue 43, pp. 144-62.

Singer, E. 1993. "Informed Consent and Survey Response: A Summary of the Empirical Literature," Journal of Official Statistics 9, pp. 361-75.

Singer, E., Schaeffer, N.C., and Raghunathan, T. 1997. "Public attitudes toward data sharing by federal agencies." International Journal of Public Opinion Research, 9, pp. 277-285

Singer, E., Von Thurn, D. and Miller, E. 1995. "Confidentiality Assurances and Response: A Quantitative Review of the Experimental Literature," Public Opinion Quarterly 59, pp. 66-77.

Trott, E. 1995. " Improving Data Quality Through Increased Data Sharing," in Federal Committee on Statistical Methodology, Seminar on New Directions in Statistical Methodology, Statistical Policy Working Paper 23, Washington D.C.: Statistical Policy Office, Office of Management and Budget, pp. 531-39.

Vacca, E. 1990. "1987 Census of Agriculture Advertising and Response Behavior Study," 1990 Annual Research Conference Proceedings, U.S. Department of Commerce, March 18-21, pp. 783-99.

Zeisset, P., Mesenbourg, T., and Marske, R., 1990. "1987 Economic Censuses Advertising and Response Behavior Study," 1990 Annual Research Conference Proceedings, U.S. Department of Commerce, March 18-21, pp. 800-828.