I have a piece in this morning’s Arkansas Democrat Gazette arguing that consolidating school districts in Arkansas to 75 countywide school districts is not a promising reform strategy. A number of state officials as well as the Dem Gaz have floated the idea of cutting the number of districts to less than one-third of the current number as a way of saving superintendent and football coach salaries while improving the capacity of high schools to offer state-required courses. I argue that the salary savings will be few, there are better ways to help high schools offer courses (such as with distance ed), and student achievement tends to suffer in larger schools and school districts.
Now, this doesn’t mean that reconfiguring larger urban high schools into “small” schools within a school, as the Gates Foundation once pushed, is likely to produce much of an improvement either. The benefits of smaller schools and school districts may be related to the tighter connection they have to their communities and the more competitive market provided by having more districts. Simply breaking up big high schools may not better connect schools to communities or create more competitive pressure.
Being able to choose among schools within a district is like being able to choose among the menu items at McDonalds. It’s nice that you could choose the Filet-O-Fish if you prefer to eat fish, but there is no change in competitive pressure from adding that menu item — all of the money still ends up in the same place. The same is true for choice within school districts — all of the money still stays with the school district, so their motivation is not significantly altered by your choice among their schools. We should only expect significant competitive pressure when money leaves one organization and enters another as a result of consumer choice. School districts are the main organizational unit of education funding.
Jay P. Greene's Blog 
The problem with size in schools and/or districts is that the management model doesn’t scale up (or down) well. When you change the size, you have to change management structure just as you have to change teaching styles when you change the size of the classroom. Size shouldn’t be the arguable point.
I agree, Jim. At the risk of sounding off-color, size is not in the issue.
Years ago, I investigated this question. NCES does not make available to unaffiliated researchers test score data on individual students, schools, or school districts, so I used three indirect measures of district size: the State-level mean district size, the fraction of total K-12 students enrollment assigned to districts over 20,000 enrollment (or 15,000 enrollment, depending on which year of the Digest of Education Statistics you use) , and the fractiion of total K-12 enrollment assigned to one or another of the nation’s top 130 largest school districts. I then used the EXCEL correlation function to compute the coefficient of correlation between these measures and State-level NAEP 4th and 8th grade Reading and Math scores, from the years 1990, 1992, 1994 (Reading only), 1996 and 2000. I used composite scores, Numbers and Operations subtest scores, Algebra and Functions subtest scores. I used percentile scores, proficiency scores, mean scores and mean scores by parents’ race and level of education. Across the board (with a couple of instructive exceptions), the correlation (size, score) is negative, usually in the range of -0.20 to -0.40.
The correlation ($/pupil, enrollment) is positive, in the Digest table “Selected characteristics of school districts over 20,000” (or 15,000 depending on which year of the Digest you use). Across States, the correlation (mean district size, $/pupil) is positive.
Per pupil costs rise and overall system performance falls as districts increase in size.