validity of the entire data set produced. It is to these issues that we 

 now briefly turn. 



THE ISSUE OF EXPERIMENTAL DESIGN 



In social science research there are basically two designs: (1) 

 cross-sectional, and (2) longitudinal (Zetterberg, 1965). The former is 

 characteristic of most survey research. The latter is represented by 

 the panel or time-series approach. Symbolically, the cross-sectional 

 design would appear as follows: 



ti ti 



Xi Yi 

 X2 Y2 

 X3 Y3 



X 



n 



where we measure a sample of n units at time t, with regard to 

 variables X and Y. The longitudinal design, on the other hand, can be 

 depicted as follows: 



ti t2 t2 



Xi Yi Xr Yf 



X2 Y2 X2 Y2 

 X3 Y3 X3 Y3 



• • • ■ 



Xn Yn An !„ 



where we measure a sample of n units at times ti and t2 with respect 

 to X and Y. 



The major difference between a cross-sectional and a longitud- 

 inal design is the introduction of time as a key factor. A dynamic test 

 of an hypothesis in the cross-sectional framework cannot be done; 

 this research design provides a test of the posited relationship Y = f (X). 



The longitudinal design, on the other hand, incorporates and 

 extends this process. Not only can one assess Y = f (X) but one can 

 now compare the "n"units at two points in time, introducing a 

 dynamic aspect to the study ( A Y = A f (X)). Because it is conceivable 

 that the former test might lead to the acceptance of the original 

 hypothesis, while the latter might lead to its rejection, longitudinal 

 designs are seen as more sensitive than cross-sectional designs. 



In order to evaluate the effectiveness of the CETA experience on 

 program enrollees, it would clearly be necessary to use a longitudi- 

 nal framework. This would allow for the monitoring of key factors 

 thought to influence the degree of success of the program (as 



