Simple random sampling in a wage and hour case

An article you may find useful:

http://knol.google.com/k/dwight-steward/simple-random-sampling/15hmam9ln4cjr/5#

Salary discrimination allegations and data

In contrast to cases involving hiring discrimination, employment cases involving discriminatory salary allegations tend to present relatively few special data issues for the statistical expert. Generally, employers tend to keep data records on compensation issues that are sufficient for most statistical employment analyses. In these types of cases, the statistical expert will generally be most concerned with the specification of the underlying compensation process.

For example, in many of the cases both the plaintiff and defendant statistical experts will utilize the same employer salary and generally use the same class of statistical methodologies. The differences between the opposing statistical experts opinions frequently revolves around the specification of the compensation model and the employment actors included in the model.

Hiring discrimination and data

Similar to cases involving discriminatory pay allegations, discriminatory employment termination type cases present relatively few data related issues for the statistical expert. Like pay cases, most of the differences revolve around the specification of the statistical model of the employer’s practices. However, determining which protected and non-protected group members were actually eligible for a given reduction in force action does tend to produce differing opinions

Statistics Saturday: Wage and Hour Statistics Tools

Sampling Error, n. the difference between the
true, and generally unknown, value of a statistic,
such as the mean average salary, of a parent
population and the value that is estimated from a
sample.

Discussion:

Sampling error in commonly presented in national
surveys. For instance, a survey may report that
27%, with a sampling error of +/-3.5%, of the
employees were interested in applying for a given
job.

The sampling error in this example would
mean that it can be expected that had the entire
population been asked the question, between
23.5% and 30.5% of the employees would have
reported that they were interested in the job in
question.

Discriminatory hiring allegations: Special Concerns

Statistical analyses involving hiring allegations tend to present a number of special issues. First, it is not uncommon for organizations to keep only cursory records on individuals who apply for a given job position. Consequently, special attention must be dedicated to the construction of the relevant applicant pool in these types of employment cases. In some instances in hiring analyses, the employer level hiring data is augmented with labor market availability information from government and other publicly available sources to provide a more probative analysis.

Second, in addition to the data issues in hiring cases, the definition of the relevant pool of qualified applicants is typically an issue of concern for the statistical expert. Generally speaking, it is important for the expert to ensure that the hiring analysis compares individuals who, but for their protected class status, would have a comparable probability of being hired. For example, in a case involving allegations of gender discrimination in faculty hiring at a university, factors such as the individual’s prior research and teaching experience would be factors to potentially consider in the construction of the qualified applicant pool.

Third, it is important to consider each of the stages that comprise a typical employer’s hiring process as well as the different factors that may effect the individual’s decisions at each stage of the hiring process. For example, many employers’ hiring processes are comprised of distinct stages that may include the application stage, interview stage, testing stage, and ultimately the selection stage. It is important to clearly define the hiring stages, as well as to determine the number of applicants in both the protected and non-protected group who are not only qualified but are truly interested, potential job candidates for the given position.

Using statistical evidence in employment cases (Part III)

(2) after accounting for relevant employment factors, does the individual’s protected class status remain a statistically important factor?

This question is typically used by the statistical expert when assessing the importance of the employee’s protected group status in the selection or pay process when there are observable employment related differences among the individuals being analyzed. For instance, it is generally recognized that an individual employee’s salary will tend to be related to factors such as the type of job performed, seniority, education, as well as company specific factors such as time in salary grade.

In these types of discrimination analyses, the expert will develop statistical models, such as linear regression models, to account for the effect that these employment factors and the protected group status collectively have on the selection or compensation process. In these types of models, a finding that the individual’s protected group member status remains a statistically important factor even after accounting for the observable employment related differences among the individuals, is typically viewed as suggestive of a nonneutral employment process.

Hypothesis testing in employment discrimination cases (part I)

Generally, in an employment discrimination lawsuit, the statistical expert will conduct statistical tests that address one or both of the following employment related questions or hypotheses:

(1) if there was no discrimination against the protected group members, what would be the probability of observing the employment related disparity by random chance alone?

(2) after accounting for relevant employment factors, does the individual’s protected class status remain a statistically important factor?

Collecting data in discrimination cases

Regardless of who the attorney is representing, it is important to provide the statistical expert with employment information that completely describes the employment processes that are being studied. Generally this will involve providing the statistical expert not only the hardcopy or electronic employment data that describes the statistical disparity, but in addition, the background information on the employer and the employer’s practices.

Typically, the background information will include information from sources such as electronic databases provided by the defendant, employer handbooks, written descriptions of the relevant selection processes, and depositions of human resource personnel and other key decision makers. If at all possible, engaging the statistical expert early in the discovery process typically will allow the expert to more adequately prepare a listing of the specific information needed in the employment analysis.

In most employment analyses the statistical expert will at a minimum require the following information about the defendant’s employment processes.

A. Employee level information. The employee level information required by the statistical expert includes not only the demographic and employment information, such as date of hire, salary grade, etc., for the plaintiff but also for all the employees in the organization being analyzed. This information will allow the statistical expert to construct comparison pools of ‘similarly situated’ employees.

B. Employer practices information. This type of information includes information about the factors that are incorporated into the employer’s selection or compensation processes. For instance, in cases involving discrimination in employee terminations, it is important to determine the specific formula or individual factors that were considered by the employer in the relevant reduction in force action.

C. Company specific information. This information generally includes company specific factors that describe the organizational differences between different divisions within the relevant analysis unit. For example, in some companies that closely tie financial performance to employee salaries, it is not uncommon to observe higher average pay levels in division that generate higher levels of revenue for the company. For larger organizations, the pay or salary grade structure for the company is also important information for the statistical expert.

Stats Saturday: Wage and hour terms to know

Stratified Sampling, n. a method of statistical
sampling that draws sub-samples from
different sections, or strata, of the overall data
population.

Discussion:

Stratified sampling routines are used in
employment settings when there are important
differences between different groups of
employees that are being surveyed.. For
example, in a survey of off-the-clock work,
workers at different locations and different
supervisors may have different work culture
that make it less likely