8 



through real estate and occupation taxes. Real estate remains the 

 chief local tax source for this state. 2 



The coefficient E^ is positive, greater than one, and significant 

 in all three regressions. However, there is considerable variation in 

 coefficient B 2 . At the MCD level the coefficient on the non : 511 taxes 

 is significant and less than negative one. At the county level, the 

 coefficient is also negative, but it is not significant and it approaches 

 zero. At the regional level, the coefficient reverses its sign, is much 

 greater than one, and remains insignificant. Again, conclusions 

 which are supported at one level of aggregation are clearly unsup- 

 portable at other levels. 



This paper has provided an empirical demonstration of some of 

 the problems inherent to aggregation which have been discussed 

 elsewhere. We have observed cases where R 2 falls with increased 

 aggregation, R 2 increases with increased aggregation, and R 2 

 remains relatively constant with increased aggregation. More sig- 

 nificantly, the coefficients have switched signs and magnitudes, in 

 some cases they have lost statistical significance, and in one case we 

 have seen a sign switch direction while retaining significance 

 (although at a lower level) as aggregation increased. 



One of the questions motivating this exercise is how does one 

 assimilate "community" research conducted at various levels of 

 aggregation. Findings of significance at one level need not general- 

 ize to the "community." Results with high predictive power may 

 obscure "community" realities, and the reverse may also be true. 

 Situations in which no constructive results emerge also need not 

 generalize to the "community." 



As pointed out by Blalock (1979) in his presidential address to 

 the American Sociological Association, this problem is endemic to 

 social science research. It is of particular importance to the com- 

 munity researcher, however, because of the lack of consensus sur- 

 rounding the definition of community. In instances where aggrega- 

 tion of data has occurred, due to decision criteria established by 

 others, the researcher needs to be aware of the limited generalizabil- 

 ity of his/her results. 



2 While we use both 51 1 and non-51 1 tax generated revenue in our models, we do not 

 mean to imply that these are the only categories of tax sources. Indeed, non-tax 

 revenues accounted for nearly 41% of all revenue generated in 1970. The source of this 

 revenue includes dollars generated through public service enterprises, water supply 

 and sewer charges, state and federal grants, licenses, permits, and fines. 



