Converting Previous Election Results to Current Geography Through Matching Elections results can be associated with current geography in different ways. A general discussion of how the Statewide Database handles geographical assignment of election results is given in ``Creating California's Official Redistricting Database'' (http://swdb.berkeley.edu/d10/Creating%20CA%20Official%20Redistricting%20Database.pdf), and it is assumed that the reader of this documentation is familar with the techniques described in that document. The method described here uses the technologies and techniques described in the ``Creating'' document, and extends them to allow a rapid and relatively accurate conversion of previous election results to any current geography. As described in the ``Creating'' document above, each individual voter is given a behavioral propensity of voting for a particular candidate or proposition. These propensities were then used in the creation of the unified dataset that was released in April of 2011. While calculation of these propensities and the subsequent merger to a common format are a time-consuming process, once calculated, these propensities can be used to provide conversion of the election results for which they were calculated to current electoral geography. This is done in two complementary ways. First, the current registered voters are matched to the registered voters from the previous election those voters behavioral propsensities are associated with the registration precinct of the current registered voter. Second, for those registered voters that cannot be matched across the two registered voter lists, a geographic conversion is created using the matched registered voters, and election results that have not been assigned are distributed through these conversions. An example is given for Alpine County, allocating g10 election results to the p12 registration precinct geography for Proposition 19. Below, rgprec refers to the p12 registration precinct and srprec refers to the g10 sr precinct (see the "Creating" document for a description of these precincts and a description of how the propensities are derived). Number of registered voters g10: 781 Number of registered voters p12: 818 Number of voters who voted in g10: 553 Number of voters who voted in g10 matched to p12 registration: 486 Voting propensities are of the form (this is for g10 sr precinct 1) PR_19_N PR_19_Y left 0.275958 0.72355 right 0.618946 0.37944 ind 0.152179 0.84782 where the groups are an assignment of voters on the basis of declared partisan affiliation (see the ``Creating'' documentation). A voter who voted in the g10 election and who match a registered voter in the p12 election are now used to calulate "matched" voting results by p12 rg precinct, by summing the propensities of these voters: rgprec PR_19_N PR_19_Y 11 43.69 65.12 21 46.63 62.72 31 17.49 27.05 41 38.44 67.81 51 46.38 65.86 Totals 192.63 288.55 Now, to assign voting results for voters in sr precincts in g10 who are not matched to p12 voters, a conversion file is created between the p12 rg precincts and the g10 sr precincts (using all registered voters who match between the two elections, not just those who voted). This conversion can be displayed as follows: rg prec tot reg tot reg sr prec tot reg sr prec tot reg matched 1 matched 2 matched ---------------------------------------------------------------- 11 139 109 1 114 2 7 21 142 111 2 100 5 23 31 107 45 3 82 5 10 41 260 108 4 142 51 170 113 2 1 5 145 To allocate g10 results, "reverse" this mapping (that is, go from g10 sr precinct to p12 rg precinct). sr prec rg prec tot reg rg prec tot reg rg prec tot reg 1 matched 2 matched 3 matched ---------------------------------------------------------------- 1 11 114 2 11 7 21 100 51 1 3 31 82 4 41 142 5 21 23 31 10 51 145 This then allows proportional allocation of sr precinct voting results to p12 rg precincts. Since the voters in the g10 srprecs who are not matched to the p12 rg voters are known, they can be displayed by group. 1 ind 4 left 6 right 7 2 ind 4 left 8 right 9 3 ind 1 left 2 right 1 4 ind 5 left 5 right 6 5 ind 1 left 6 right 2 The above propensities can then be applied to obtain vote totals by srprec. srprec PR_19_N PR_19_Y 1 6.60 10.39 2 9.61 11.07 3 1.35 2.60 4 5.61 10.17 5 2.97 5.94 Totals 26.14 40.17 The voting totals by srprec are also known: srprec PR_19_N PR_19_Y 1 47 74 2 54 65 3 16 25 4 45 80 5 62 91 Totals 224 335 Also known are the "matched" totals by g10 sr precinct (note the county totals match those for the p12 rg precincts given above, though the individual table entries won't generally agree as precinct boundaries shift between elections and sr precincts and rg precincts are often not directly comparable). srprec PR_19_N PR_19_Y 1 39.98 61.93 2 43.50 52.86 3 12.93 23.65 4 38.44 67.81 5 57.78 82.31 Totals 192.63 288.55 Adjustments are then made to the propensities to obtain equality of the voting results to the sum of the matched and unmatched voting results. Finally, the adjusted unmatched totals are mapped back to the p12 rg precincts and, together with the matched totals, give a description of the g10 voting results by p12 rg precinct. Matched Unmatched SOV Totals rgprec PR_19_N PR_19_Y PR_19_N PR_19_Y PR_19_N PR_19_Y 11 43.69 65.12 7.70 12.86 21 46.63 62.72 10.27 12.37 31 17.49 27.05 3.31 1.84 41 38.44 67.81 6.56 12.19 51 46.38 65.86 3.54 7.20 Totals 192.63 288.55 31.37 46.45 224 335 The SOV totals for p12 rg precincts cannot be known but only estimated, but the estimated totals do sum to the county totals.