substantially different from the ACS rates in many states. In the most populous state, California, the state reports yielded an immigrant student estimate of 4.1 percent of all students in school year 2007-2008 while the ACS estimate was 1.4 percent. In most, but not all states, the state-reported rates were higher than the ACS rates.

Table 6-12 also summarizes the strength of the association between ACS and state-provided rates within each state, corrected for overfitting due to sampling error. With some exceptions, these correlations tend to be generally quite high, with half of the 30 states included showing adjusted correlations of higher than 0.90. This finding suggests that the measures are usually fairly consistent within each state, holding constant the state procedures and to some extent the immigration patterns (to the extent that they are more consistent within than between states). However, the correlations are considerably weaker in some states, perhaps providing evidence of inconsistent collection of immigration data or of varying patterns of immigration that affect consistency of reporting.

We also tested the relationship at the school district level between ACS estimates and state-provided estimates of rate of immigrant children among public school enrollees. As shown in Table 6-13, the results are mixed, with some states showing a very good consistency between the ACS and state-provided numbers for immigrants, and other states showing a very weak relationship between the series. When compared with the results of this test for the ELL estimates and counts in Chapter 5, these findings suggest that there are perhaps systemic differences between the ACS and state-provided counts at the school district level. The results suggest the possibility of less consistent procedures and criteria within many states than was observed with the within-state counts of ELL students, an indication that caution should be exercised in using the state-provided counts of immigrant children.