freedom, there is seen a pattern of increasing R 2 with the level of 

 aggregation. This is an expected pattern but not universal. With the 

 MCD equation both variables are significant and R 2 indicates that 

 there are a number of influences that have not been controlled. At 

 the regional level, R 2 is so high that there do not appear to be many 

 influences left to control for, and, indeed, changes in earned income 

 taxes account for very little of the variance of the dependent varia- 

 ble. It is debatable whether the regional equation explains more 

 than the MCD equation or simply hides more. 



The second model investigated represents a different but related 

 attempt at accounting for the observed level of 1974 per capita tax 

 revenue. Two variables were again selected: (1) the change in earned 

 income taxes (defined and utilized as in Model 1), and (2) the change 

 in total Act 511 tax dollars collected. The results are presented in 

 Table 2. Each row in this table includes the constant («*), the esti- 

 mated coefficient on the change in earned income taxes (Bj), the 

 estimated coefficient on the change in Act 511 collections (B 2 ), and 

 the summary measures as before. 



Table 2. Summary of Regression at Three Levels 

 of Aggregation for Equation 2. 



OC Bi B 2 R 2 Adj. n 



Aggregation 



Minor Civil Division 749 

 County 458 



Region 193 



♦Significant at .001. 



'Numbers in parentheses are simple t statistics. 



At the MCD level, changes in earned income taxes have a nega- 

 tive impact, while changes in Act 511 collections have a positive 

 impact. Both of these are significant. One possible interpretation is 

 that, as income grows, localities become less reliant absolutely on 

 per capita taxation because other collections are positively related 

 to income. Moreover, areas that have changed the level of Act 511 



