ter income earned by each employee—this is similar to the U.S. form W2. At the same time, they must indicate who was employed on a specific day (typically in November). Furthermore, for firms that maintain multiple work sites, employers indicate in the same form the physical location at which each employee worked throughout the year.
Government agencies are allowed to share data only to a limited extent and, even then, it must be permitted in the act that authorizes collection of specific data. However, Statistics Denmark can by law request all data, including the person and firm IDs, from each collecting agency, as well as from private sources. Statistics Denmark is thereby allowed to fully exploit the common ID numbering system and use all data in the production of statistics and for research. Trust and reliance in the system are so broad that a standard census has not been conducted since 1970. One of the conditions of use by Statistics Denmark is that no individual or firm can be identified. Researchers desiring access to confidential microdata that are allowed work inside the agency are subject to the same nondisclosure requirements as regular employees. For the past five years or so, it has also been possible for researchers to obtain admission to approved data sets from their own workplaces through the Internet (http://www.dst.dk/HomeUK/ForSale/Research.aspx).
Although the approaches described here may be more problematic to implement for an economy as large and complex as the United States (e.g., where unregistered and illegal employment is perhaps more widespread), universal coverage registers create interesting possibilities that do not exist with other systems. First, survey samples of persons and businesses can be drawn from the universal registers. This increases the quality of sampling frames and saves respondents time because they do not have to provide information already embedded in the registers (number of employees, industry, gender, age, education, children, etc.). Second, survey response data can be easily merged with register information. For example, responses from a survey of job and life satisfaction, in which only one or a few employees in each firm are sampled, have been linked with register information on all the coworkers in the firm. This has generated information on how the composition of workers, their wages, and the hierarchical structure of the firm may affect responses from the sampled person. Another example is analysis of the relation between human capital formation of employees and the bottom line of firms. A different class of examples can be found in medical research, where groups of patients can be traced back in time with controls for work- or residence-related exposure applied. An example here is work on the long-term effects of occupational hazards.