This essay updates my previous essay: "Allocation of NCA slots: Is there a rationale to the 2003 Allocations?" The updates are based largely on Prof. Watson's correction of the assumptions in that prior article. They also reflect continued exploration of the tacit rationale for the 2003 slot allocations that were explored in that essay. In particular, it makes use of a more "finished" spreadsheet model in which the assumptions can be quickly changed and their effects on two alternative allocation methods can be viewed. The goal of the exercise remains unchanged. This paper, like its predecessor, seeks to recreate algorithmically the tacit algorithm that emerged discursively over the course of prior conventions. It is hoped the result will prove helpful in discussions of what assumptions should be used in the future. The more "finished" Excel format spreadsheet used in this paper is available online for download by anyone who would like to view it and play with its assumptions, which are as follows:

1. that affiliated organizations are allocated slots independently of the membership-based algorithm used for the 2004 slot allocations. A prior assumption, that affiliate organizations pay for their slots, has turned out to be erroneous. It remains, however, that these allocations are essentially unchanged from 2003 to 2004. It is assumed, in these alternative models and therefore shouldn't have their slot allocations altered
   
2. that there is a minimum number of slots that any division should get before allocations were made based on membership. This assumption parallels the algorithm used in determining the number of seats individual states have in congress, where every state gets two senators regardless of their population and a variable number of representatives based on that population. This seems sensible if only because at least one slot should be allocated to each division or caucus for its business meeting. In the prior paper I found that a baseline of 5 slots most closely approximated the 2003 allocations. The explorations in this paper generally suggest that that baseline is either 5 or 6 slots depending on the values for other assumptions. The spreadsheet-based values should in the table below are based on an assumption of a 6 slot baseline, as this value maximizes the fit with the 2003 allocations given the other assumptions used in the table.
3. that no division can reasonably be expected to make good use of many more slots than their are time periods available in which slots can be scheduled. Specifically, given that there are a maximum of 22 to 25 slots at a convention, with 2 or 3 of those are typically given over to the business meeting and keynote(s), it should be expected that a division cannot make use of more than 20 to 23 slots without competing against itself. Whether or not this is a reasonable assumption, it remains that the maximum divisional allocation of 27 slots in 2003 is consistent with it. Hence the spreadsheet-based values reported in this paper are based on an assumption of a maximum of 27 slots per division, an increase from the 25 used in this last paper, as this value matches the maximum value used in 2003. It should be noted, however, that the fit with the 2003 allocations is optimized with a maximum of 22 slots.
4. that the slot requirements of caucuses should be recognized as being different than slot allocations of divisions. The NCA constitution assigns to divisions responsibility for "topics of substantive study" and assigns to caucuses representation of the "interests of members in specific demographic or socially-defined groups seeking to realize the objectives specified in the National Communication Association's Affirmative Action Statement." While the latter mission is important, it can be reasonably argued that it does not merit the same level of slot assignment the former does. Indeed, it might reasonably be argued that a large caucus membership is a measure of a groups success in closing the opportunity gaps that affirmative action attempts to close such that caucus slot assignments should be inversely proportional to membership. In the prior paper an equal number of slots were assigned to each caucus. This paper caps caucus assignments at 18, which is consistent with the maximum number of slots allocated to caucuses in 2003. It should be noted, however, that the fit with the 2003 allotments is optimized when caucuses are granted a maximum of 20 slots.
   

Two algorithms are reported here. The first, which is identical to the "5+ slot allocation method" of the previous essay in every essential but the minimum number of slots assumed, makes the first two assumptions above. The second, which differs from the "25 Max slot allocation method" of the previous essay in its assignment of a maximum number of caucus slots rather than a fixed allocation of caucus slots and in the maximum number of slots assigned to divisions. There are no substantial changes in the correlations associated with these models from the prior paper:

• The current 2004 slot allocation correlates .76 with the 2003 allocations. This differs from the .73 correlation reported in the prior paper because affiliates had been excluded from the regression in the prior paper (an error).
• The minimum sessions algorithm correlates .83 with the 2003 allocations and .97 with the 2004 allocations. Taken in combination, these correlations suggest that moderate changes in the assumptions in the slot allocation algorithm can substantially recapture one of the tacit assumptions in 2003 without substantially reducing the effectiveness of the 2004 algorithm in allocating more slots to larger divisions.
• The maximum sessions algorithm correlates .89 with the 2003 allocations and .90 with the 2004 allocations. This is a fairly interesting outcome. The maximum sessions algorithm is intended to approximate the 2003 allocations as closely as possible. Its capture of 78% of the variance in the 2003 allocations is an indication that its assumptions provide a fairly good approximation of the unstated, but discursively arrived at, rationale behind those allocations. Its success in capturing 81% of the variance in the 2004 algorithm is an indication of just how important division size has been in the allocation of slots..

The purpose of this essay is not to suggest that any of these algorithms is better than the others, but rather to see if the old informal allocation method had a rationale that could be modeled algorithmically and to make that algorithm more accessible to readers. The previous paper established that there was a rationale to the 2003 Allocations. It is hoped that this paper, and its associate spreadsheet, will be of use in mapping future NCA convention slot allocations.