In the U.S., it may be more difficult to obtain good administrative data for an entire population than in many other countries. Partial data may be useful, but some may try to use it for substitution-type imputation, and that seems problematic. The real value in administrative data is in its use as auxiliary data for model-assisted design-based estimation, such as found in Särndal, C.-E., Swensson, B. and Wretman, J. (1992), Model Assisted Survey Sampling, Springer-Verlag, or as noted in Särndal, C.-E., and Lundström, S., (2005), Estimation in Surveys with Nonresponse, Wiley.
Also, if administrative data on an entire population can be obtained, it may be used as regressor data in strictly model-based estimation.
For official statistics, particularly for establishment surveys, one may have an infrequent census that could be used as auxiliary/regressor data for a more frequent sample on the same data items. But if such auxiliary or regressor data are not available, administrative data could be very useful. From Särndal, and Lundström (2005), it seems that such administrative data may be more commonly available in Europe than in the US.
A major use of auxiliary/regressor data may be in various uses of ratios in estimation, as shown in the paper at the first link below.
What sources of administrative data have you used or noted for any such survey? Where might such data be obtained? How did you find it? This would be highly dependent on the subject matter, but may stimulate/inspire ideas in others as to how to obtain administrative data for their use. Such efficient use of available information may be very helpful for many organizations.
Article Use of Ratios for Estimation of Official Statistics at a Sta...
Article Book Review for Journal of Official Statistics on Estimation...
Collecting auxiliary variables on respondents and nonrespondents to guide attempts to balance response rates across key subgroups is wise. AD can be thought of as encompassing all information not obtained directly from the interview of a respondent or case - specific informant. First, it starts with information from the sample frame itself. What is available varies greatly. In many samples of special populations there are often sample frames not only rich in information, but having abundant data closely related to the subject of the survey being conducted.
Matters as gaining more familiarity with many auxiliary databases, learning about using GPS and GIS, better mastery of multi- level, geographical data, and improving record linkage procedures. In brief, it means that the repertoire of skills and expertise necessary for conducting survey research needs to expand.
Adding more AD increases the likelihood of deductive disclosure and thus potentially undermines confidentiality. Files with extensive AD would either have to be restricted, sensitive - data files and/or the AD would have to be partly masked to eliminate the possibility of deductive disclosure. Data files can be thought of as falling along a sensitivity continuum from public -use files that have been designed to ensure that respondent identity cannot be broached via deductive disclosure or other means to in-house files with explicit respondent identifiers. In between these are restricted-access data files which contain more detailed information than the public files, but lack explicit identifiers. These may be accessible to researchers via sensitive-data access agreements and / or via data enclaves.
http://www.norc.org/PDFs/ARC/CSPS/midareport615.pdf
Regarding official use, the major problem with obtaining administrative records in the US appears to be that in the US, the various agencies are greatly restricted from sharing information with each other, and even within an agency. In contrast, I've heard that is not such a problem for, say, Statistics Canada.
Comments?
THis is just to say that the UK is making quite large scale investments in admin data through national centres
http://www.esrc.ac.uk/news-and-events/press-releases/28673/the-big-data-family-is-borndavid-willetts-mp-announces-the-esrc-big-data-network.aspx
New Zealand is also investing in AD, partly to reduce costs, and partly to reduce respondent burden (both more important in a small country where sample sizes are relatively greater than for a big country), and also for better SAE. Privacy issues are well known as well as the fact that AD is not designed for statistical purposes. Sometimes an organisation can be persuaded to collect a little more information.
For privacy, linked data is not sent back to the institution supplyiing it.
Here's one project that makes extensive use of AD: http://www.stats.govt.nz/leed
The Census Bureau Center for Economic Studies (http://www.census.gov/ces/) has extensive data on establishments. Some of the data are restricted use, so there is a process to obtain access to the data. CES also has public-use data on US businesses.
- Tax Authority Records
- Federal Labour Office
- Chamber of Commerce and trade
- Handicrafts Chamber etc.,
http://www.geostat.ge/cms/site_images/_files/georgian/methodology/business/15.%20Use%20of%20administrative%20sources%20for%20business%20statistics%20purposes%20Handbook%20on%20good%20practices%20%281999%20Edition,%20Eurostat%29.pdf
- Badges
- Science topic
- Similar topics
- Statistics
- Survey