Abstract
The purpose of this dissertation is to characterize the job-separation behavior of new entry-level employees. The specific case considered is that of enlistees in the U.S. Air Force, from September 1974 to August 1977. Two general sets of characteristics will be used to explain separations, or first-year losses. Person-specific characteristics explain differences in behavior across individuals who face the same external conditions but respond to those conditions differently. General economic conditions and the environmental constraints faced by the enlistees (entry-level workers) are the second set of explanatory characteristics. These characteristics will explain differences in first-year loss rates through the business cycle and as the Air Force alters its decision rules. The econometric model used to organize the data is a binary logit model. This research suggested that three types of variables are important in explaining the first year loss rates. The first, a proxy for environmental constraints imposed by the Air Force, was the quarterly mandated force level and its associated growth. These two variables were the most dominant encountered in this research. This suggests that the decision rules and protocol of the Air Force are extremely important determinants of first-year loss behavior. The second result is that increases in the military compensation tend to increase first-year loss rates. Although this is inconsistent with nonstochastic economic theory, stochastic theories can be developed that allow this type of behavior. The third result is to confirm the outcomes of earlier research where intelligence and education measures of the individuals, prior to enlistment, affect first-year loss rates. Those, who score higher on an entrance aptitude test and/or have graduated from high school tend to have lower first-year loss rates.
Worley, Lawrence Franklin (1986). An econometric analysis of early employment losses : the case of the U.S. Air Force. Texas A&M University. Texas A&M University. Libraries. Available electronically from
https : / /hdl .handle .net /1969 .1 /DISSERTATIONS -16955.