130 



Journal of the Kentucky Academy of Science 66(2) 



ogy transfer, intellectual property, and royalty 

 income are among the important issues di- 

 rectly linking the development of intellectual 

 capital to the state's economic growth and 

 viability. 



METHODS 



The Kentucky baccalaureate origins of doc- 

 torate recipients in the biological sciences, 

 chemistry, and physics are derived from data 

 provided by the Survey of Earned Doctorates 

 (SED) and entered into WebCASPAR (Com- 

 puter-Aided Science Policy Analysis and Re- 

 search). The SED is conducted annually by 

 the University of Chicago National Opinion 

 Research Center for the Federal sponsors of 

 the survey: the National Science Foundation 

 (NSF), the National Institutes of Health 

 (NIH), the U.S. Department of Education 

 (USED), the National Endowment for the Hu- 

 manities (NEH), the U.S. Department of Ag- 

 riculture (USDA), and the National Aeronau- 

 tic and Space Administration (NASA). The 

 SED does not report professional degrees such 

 as the D.D.S., D.V.M., J.D, M.D., and O.D. 

 Information from this survey becomes part of 

 the Doctorate Records File (DRF) and con- 

 tains data on all earned doctorates granted by 

 regionally accredited U.S. universities in all 

 fields from 1920 to the present. 



WebCASPAR is a database system (NSF- 

 WebCASPAR 2005) available on the World 

 Wide Web and containing information about 

 academic science and engineering resources. 

 Information from academic surveys of the Di- 

 vision of Science Resources Statistics (SRS) 

 of the National Science Foundation and infor- 

 mation from a variety of other sources, in- 

 cluding but not limited to the SED, are in- 

 cluded in the database. WebCASPAR 

 currently can provide critical demographic 

 data through access to most DRF doctoral rec- 

 ords in the biological sciences, chemistry, and 

 physics from 1966 through 2003. 



Data for the SED are collected directly 

 from individual doctorate recipients. The 

 questionnaire is distributed through the co- 

 operation of graduate deans to those who are 

 completing their doctorates. The data for a 

 given year include all doctorates awarded in 

 the 1 2-month period ending on 30 June. Over 

 the past decade, the self-report response rates 

 to the SED questionnaire varied from a high 



of 94.7% in 1993 to a low of 91.2% in 2002. 

 The response rate for a specific year may vary 

 slightly over time as late responses are added. 

 To minimize the effect of late responses, we 

 used 2002 as the last year for our data survey. 



Since, "differences in survey and item re- 

 sponse rates from year to year can produce 

 numerical fluctuations that are unrelated to 

 true trends" (NSF Science Resources Statis- 

 tics 2005), we used a 25-year time line (1978- 

 2002) to minimize the effect of such fluctua- 

 tions. Through this paradigm we addressed the 

 stated concern that "increasing or decreasing 

 numbers in a citizenship or racial/ethnic group 

 . . . reflect to some degree any upward or 

 downward change in both overall survey re- 

 sponse and item response" (NSF Science Re- 

 sources Statistics 2005). 



Until 1990, SED records with missing gen- 

 der were imputed using the first name of the 

 recipient only if the first name could be pos- 

 itively identified as male or female. Records 

 that could not be identified by first name were 

 set to male. In 1990, follow-up procedures 

 were implemented for cases still missing after 

 imputation. Records for which gender still 

 could not be identified in the follow-up were 

 set to male. In 1997, it was decided to dis- 

 continue this practice. Therefore, data with 

 unknown gender occur in 1988 and in 1990 

 and later years. For each year beginning in 

 1990, however, the response rate is greater 

 than 99.2% (Hill, S. T. 2005. Director, Doc- 

 torate Data Project, NSF Division of Science 

 Resources Statistics, pers. comm., 11 Feb.). 



As a result of new follow-up procedures 

 implemented during the 1990 survey cycle, 

 there was greater adjustment to the numbers 

 for race/ethnicity and citizenship than in ear- 

 lier years. Although these new procedures 

 have improved the completeness of the data, 

 they also created a break in trend data that 

 should be taken into consideration when an- 

 alyzing sensitive data such as race/ethnicity 

 after 1989. The effect on the trend line for 

 citizenship is extremely small, almost negli- 

 gible. The effect on the trend line for race is 

 minimal. To the extent that there has been an 

 effect, blacks tend to be slightly overrepre- 

 sented in institution-completed responses, so 

 blacks would trend toward a larger proportion 

 after the new procedures were implemented 

 (Hill, S.T. 2005. Director, Doctorate Data Proj- 



