collections are likely to adjust their per capita collections in the 

 same direction. The value of R 2 for the minor civil division regres- 

 sion is low (.209). 



Not unexpectedly, major differences appear at the remaining 

 levels of aggregation. At the county level, neither coefficient is 

 significant, although their signs are the same as they were at the 

 MCD level. Moreover, the county equation has virtually no predic- 

 tive power. At the regional level, the equation has changed consid- 

 erably. The signs on the coefficients have switched (from the direc- 

 tions of both the minor civil division and county level equations 

 respectively). Further, R 2 has rebounded to its earlier level. How- 

 ever, unlike the first equation, neither coefficient is significant at 

 the .001 level, although the t values are not small. 



Obviously these are perplexing results. Despite the condition 

 that the same data are used in all three cases, the empirical results 

 are not generalizable. There is no apparent relation at the county 

 level, and contrary relations exist at the MCD and regional levels. 



The final model investigated attempts to account for the 

 changes in total tax revenues collected during the period 1970-1974. 

 The results are presented in Table 3. This change is viewed as a 

 function of the changes in Act 511 taxes (B : ) and the changes in 

 non-Act 511 taxes (B 2 ). The latter taxes are primarily generated 



Table 3. Summary of Rgressions at Three Levels 

 of Aggregation for Equation 3. 



Level0f <X B, B 2 R 2 Adj. n 



Aggregation 



Minor Civil Division 5009 

 County -4172 



Region -7068 



*Significant at .001. 



'Numbers in parentheses are simple t statistics. 



