The dataset contains confidential computerized personnel records for all management employees of a medium-sized U. S. firm in a service industry over the years 1969-1988. The principal investigators use these data to peer inside the "black box" of the firm to explore the existence and nature of the internal labor market and the wage policy of the firm.
The researchers obtained the firm's yearend backup personnel tapes, which included current information on every managerial employee in the firm as of December 31 of each year. Each observation contains an employee 1D number, age, sex, race, education, job title, cost center description, cost center code, salary, bonus, salary grade, and performance rating. Eight job levels from entry management (level 1) to Chairman-CEO (level 8) were clearly identified from analysis of the job titles. Seventeen titles out of over three hundred constitute a large share of employment. Levels 1-4 contain the bulk of managers, as the hierarchy narrows considerably from levels 4 to 5. All salary data are in local currencies. Not all variables are available for all years or records, although on the whole the dataset is complete. There are two major exceptions:
Bonusdata cover 1981-1988. Also, titles were not coded for some new hires in the last years, though other variables were. Thus, assignment of these employees to levels was impossible in those years. This is an insignificant problem except in 1987-1988, in which roughly 10 percent of employees and half of new hires did not have title data. These missing data mean that researchers must handle inferences from title and level data in the last few years with some care. Where the data report pooled results, the project have always calculated the statistics over 1969-1985 to test for robustness of inferences.
The timing of the variables is worth noting. Salary, title, performance rating, and other variables are year-end values. The project does not know when during the year pay or title changes occurred or ratings were given, so these variables may not be exactly concurrent. In the statistical work the researchers assume that title changes, pay changes, and ratings occur simultaneously.
Finally, for entry or entrants into the dataset it is unclear in what year those in the dataset in 1969 actually entered the firm. In addition, it is not possible to tell whether new entrants in any year are new hires at the firm; they could have been promoted from clerical to management positions.
