\n
\n\n \n \n Grofman, B. N.; and Wilson, M. C.\n\n\n \n \n \n \n \n Models of inter-election change in partisan vote share.\n \n \n \n \n\n\n \n\n\n\n
J. Theoretical Politics, 34: 481–498. 2022.\n
\n\n
\n\n
\n\n
\n\n \n \n paper\n \n \n \n slides\n \n \n\n \n\n \n link\n \n \n\n bibtex\n \n\n \n \n \n abstract \n \n\n \n \n \n 10 downloads\n \n \n\n \n \n \n \n \n \n \n\n \n \n \n \n \n\n\n\n
\n
@Article{GrWi2020,\n author = {Grofman, Bernard N. and Wilson, Mark C.},\n title = {Models of inter-election change in partisan vote share},\n pages = {481--498},\n volume\t = {34},\n issue = {4},\n journal = {J. Theoretical Politics},\n abstract = {Consider an election of a given type (say a legislative or presidential\nelection) involving some set of districts for which we have data at two\ndistinct points in time, time 1 and time 2. If there are only two\npolitical parties, A and B, we may think of the overall difference in\nthe mean party vote share of party A between the two elections as the\naggregate inter-election swing between the two parties. If we know this\nmean inter-election swing, the question we seek to answer is: ``How do\nwe expect the aggregate swing to be distributed across the districts or\nstates (or smaller units such as counties) as a function of previous\nvote share (and perhaps, other factors)?''\n\nIn the electoral systems and political party literatures there have been\ntwo main answers to that question: uniform swing and proportional\nswing. Our main theoretical contributions are (a) to provide an\naxiomatic foundation for desirable properties of a model of\ninter-election changes in vote shares in a districted legislature; (b)\nto use those axioms to demonstrate why using uniform swing or\nproportional swing is a bad idea, (c) to provide a reasonably simple\nswing model that does satisfy the axioms, and (d) to show how to\nintegrate a reversion to the mean effect into models of inter-election\nswing.\n\nOur main empirical contributions address the question of why, despite\ntheir theoretical flaws, there is strong evidence that the two standard\nmodels, especially uniform swing, provide a very good fit to empirical\ndata. We show that these models can be expected to work well when (a)\nelections are close, or (b) when we restrict ourselves to data where\nswing is low, or (c) when we eliminate the cases where the model is\nmost likely to go wrong. In particular, sometimes the model tested is\nnot the standard model; in that either (c1) a piecewise or truncated\nvariant of the model is being used, or (c2) there is a correction\n(usually 75\\%) for districts that are uncontested. As we show\nempirically with data from U.S. congressional elections, either of these\ncorrections will at least marginally improve fit on one or more of five\nindicators: mistakes about directionality of change, mistakes in winner,\nestimates that are outside the [0..1] bounds, mean-square error, and\ncorrelation between actual and predicted values. We also show that our\nnew model provides an equal or better fit to U.S. congressional data\nthan the traditional models, while having much nicer axiomatic\nproperties.},\n keywords = {electoral systems},\n url_paper = {https://markcwilson.site/Research/Outputs/GrWi2020.pdf},\n url_slides = {https://www.youtube.com/watch?v=C0eIF1oAk-U},\n year = {2022},\n}\n\n
\n
\n\n\n
\n Consider an election of a given type (say a legislative or presidential election) involving some set of districts for which we have data at two distinct points in time, time 1 and time 2. If there are only two political parties, A and B, we may think of the overall difference in the mean party vote share of party A between the two elections as the aggregate inter-election swing between the two parties. If we know this mean inter-election swing, the question we seek to answer is: ``How do we expect the aggregate swing to be distributed across the districts or states (or smaller units such as counties) as a function of previous vote share (and perhaps, other factors)?'' In the electoral systems and political party literatures there have been two main answers to that question: uniform swing and proportional swing. Our main theoretical contributions are (a) to provide an axiomatic foundation for desirable properties of a model of inter-election changes in vote shares in a districted legislature; (b) to use those axioms to demonstrate why using uniform swing or proportional swing is a bad idea, (c) to provide a reasonably simple swing model that does satisfy the axioms, and (d) to show how to integrate a reversion to the mean effect into models of inter-election swing. Our main empirical contributions address the question of why, despite their theoretical flaws, there is strong evidence that the two standard models, especially uniform swing, provide a very good fit to empirical data. We show that these models can be expected to work well when (a) elections are close, or (b) when we restrict ourselves to data where swing is low, or (c) when we eliminate the cases where the model is most likely to go wrong. In particular, sometimes the model tested is not the standard model; in that either (c1) a piecewise or truncated variant of the model is being used, or (c2) there is a correction (usually 75%) for districts that are uncontested. As we show empirically with data from U.S. congressional elections, either of these corrections will at least marginally improve fit on one or more of five indicators: mistakes about directionality of change, mistakes in winner, estimates that are outside the [0..1] bounds, mean-square error, and correlation between actual and predicted values. We also show that our new model provides an equal or better fit to U.S. congressional data than the traditional models, while having much nicer axiomatic properties.\n
\n\n\n