Rolling origin
Knowing only who the parties are, a model calls 63.5% of outcomes. Knowing only that someone is the petitioner calls 64.0%. The model is fit on the prior 25 terms and scored on the next term, from 1950 through 2014. It does not report accuracy on cases it trained on. The comparison is the rule that the petitioner wins, which is right 64.0% of the time on the same 8,095 test cases.
63.5% party-type model n = 8,095
64.0% always-petitioner baseline same test cases
0.2272 Brier score, model baseline 0.2282
26,969 cases with outcome and party type
The backcast 0% 0% 50% 50% 100% 100% PREDICTED PETITIONER WIN
Download the rows (CSV)
Table bin predicted actual n (0.0, 0.4] 0.343 0.505 503 (0.4, 0.5] 0.453 0.554 884 (0.5, 0.6] 0.559 0.614 1792 (0.6, 0.7] 0.647 0.653 2721 (0.7, 0.8] 0.744 0.737 1754 (0.8, 1.0] 0.820 0.744 441
Accuracy by term 40% 50% 60% 70% 80% 1950 1960 1970 1980 1990 2000 2010 2014 Baseline Party types Download the rows (CSV)
Table term model_accuracy baseline_accuracy n 1950 0.586 0.455 99 1951 0.532 0.617 94 1952 0.591 0.470 115 1953 0.547 0.535 86 1954 0.574 0.362 94 1955 0.544 0.447 103 1956 0.661 0.354 127 1957 0.589 0.649 151 1958 0.544 0.610 136 1959 0.647 0.632 133 1960 0.570 0.548 135 1961 0.621 0.716 116 1962 0.764 0.771 157 1963 0.698 0.771 179 1964 0.679 0.709 134 1965 0.607 0.736 140 1966 0.572 0.717 145 1967 0.694 0.765 196 1968 0.757 0.707 140 1969 0.645 0.667 141 1970 0.583 0.632 144 1971 0.615 0.631 179 1972 0.737 0.672 186 1973 0.665 0.682 173 1974 0.662 0.675 154 1975 0.714 0.714 182 1976 0.615 0.648 179 1977 0.697 0.677 155 1978 0.656 0.669 160 1979 0.622 0.628 156 1980 0.662 0.682 151 1981 0.716 0.710 176 1982 0.590 0.608 166 1983 0.717 0.711 173 1984 0.613 0.613 163 1985 0.634 0.627 161 1986 0.660 0.641 156 1987 0.487 0.487 152 1988 0.543 0.517 151 1989 0.597 0.576 139 1990 0.646 0.646 127 1991 0.667 0.650 120 1992 0.619 0.619 118 1993 0.538 0.473 93 1994 0.598 0.620 92 1995 0.611 0.578 90 1996 0.688 0.667 96 1997 0.596 0.556 99 1998 0.600 0.600 90 1999 0.573 0.585 82 2000 0.563 0.632 87 2001 0.690 0.690 84 2002 0.655 0.679 84 2003 0.734 0.747 79 2004 0.700 0.675 80 2005 0.655 0.701 87 2006 0.720 0.667 75 2007 0.521 0.616 73 2008 0.651 0.735 83 2009 0.685 0.750 92 2010 0.624 0.671 85 2011 0.610 0.623 77 2012 0.544 0.608 79 2013 0.653 0.733 75 2014 0.620 0.704 71
What each added fact buys Layer Baseline Model Test cases Party types, shrunk cell rates 64% 63.5% 8,095 party types only 64.0% 63.6% 8,095 party types + issue area 64.0% 64.7% 8,095 party types + issue + lower-court direction + source 64.0% 64.7% 8,095
Win rates by party type, modern era Type As petitioner As respondent Either side, n Business 59% of 1,758 31% of 1,951 3,709 Federal government 75% of 1,616 50% of 1,546 3,162 Individual 61% of 1,534 34% of 1,360 2,894 State government 68% of 1,120 38% of 1,637 2,757 Criminal defendant or prisoner 60% of 1,322 22% of 773 2,095 Government official or employee 62% of 414 35% of 451 865 Local government 59% of 231 36% of 336 567 Nonprofit, religious or political org 67% of 266 30% of 238 504 Union 63% of 219 31% of 171 390 Tribe 68% of 44 29% of 80 124 Foreign entity 53% of 47 29% of 28 75
Matchups farthest from the baseline Petitioner Respondent Petitioner win rate n Union Federal government 42% 38 Local government Nonprofit, religious or political org 44% 16 State government Federal government 44% 63 Business Federal government 45% 365 Business Nonprofit, religious or political org 82% 34 Federal government Government official or employee 80% 54 State government Criminal defendant or prisoner 80% 470 Government official or employee Local government 50% 44 Nonprofit, religious or political org Local government 78% 37 State government State government 50% 147 Individual Federal government 51% 461 Federal government Criminal defendant or prisoner 77% 240