1 00:00:09,490 --> 00:00:11,750 In this lecture, we'll see how analytics 2 00:00:11,750 --> 00:00:15,330 can be used to predict the outcomes of cases in the United 3 00:00:15,330 --> 00:00:17,580 States Supreme Court. 4 00:00:17,580 --> 00:00:19,740 This seems like a very unconventional use 5 00:00:19,740 --> 00:00:24,060 of analytics, but in 2002 a group of political science 6 00:00:24,060 --> 00:00:28,600 and law academics decided to test if a model can do better 7 00:00:28,600 --> 00:00:30,590 than a group of experts at predicting 8 00:00:30,590 --> 00:00:33,580 the decisions of the Supreme Court. 9 00:00:33,580 --> 00:00:36,590 In this case, a very interpretable analytics method 10 00:00:36,590 --> 00:00:42,340 was used, called classification and regression trees. 11 00:00:42,340 --> 00:00:44,520 The legal system of the United States 12 00:00:44,520 --> 00:00:46,840 operates at the state level and at 13 00:00:46,840 --> 00:00:49,600 the federal or country-wide level. 14 00:00:49,600 --> 00:00:52,460 The federal level is necessary to deal with cases 15 00:00:52,460 --> 00:00:56,560 beyond the scope of state law, like disputes between states, 16 00:00:56,560 --> 00:00:59,690 and violations of federal laws. 17 00:00:59,690 --> 00:01:02,900 The federal court is divided into three levels-- 18 00:01:02,900 --> 00:01:08,590 district courts, circuit courts, and the Supreme Court. 19 00:01:08,590 --> 00:01:10,830 Cases start at the district courts, 20 00:01:10,830 --> 00:01:14,480 where an initial decision is made about the case. 21 00:01:14,480 --> 00:01:17,310 The circuit courts hear appeals from the district courts, 22 00:01:17,310 --> 00:01:20,260 and can change the decision that was made. 23 00:01:20,260 --> 00:01:22,830 The Supreme Court is the highest level 24 00:01:22,830 --> 00:01:26,120 in the American legal system and makes the final decision 25 00:01:26,120 --> 00:01:26,620 on cases. 26 00:01:29,420 --> 00:01:31,810 The Supreme Court of the United States 27 00:01:31,810 --> 00:01:35,130 consists of nine judges, or justices, 28 00:01:35,130 --> 00:01:37,580 who are appointed by the President. 29 00:01:37,580 --> 00:01:40,960 This image shows the nine Supreme Court justices 30 00:01:40,960 --> 00:01:45,420 from the time period 1994 through 2005. 31 00:01:45,420 --> 00:01:47,310 This was the longest period of time 32 00:01:47,310 --> 00:01:52,200 with the same set of justices in over 180 years. 33 00:01:52,200 --> 00:01:56,150 The people appointed as Supreme Court justices are usually 34 00:01:56,150 --> 00:01:59,460 distinguished judges, professors of law, 35 00:01:59,460 --> 00:02:02,690 or state or federal atttorneys. 36 00:02:02,690 --> 00:02:05,870 The Supreme Court of the United States, or SCOTUS, 37 00:02:05,870 --> 00:02:09,270 decides on the most difficult and controversial cases 38 00:02:09,270 --> 00:02:11,300 in the United States. 39 00:02:11,300 --> 00:02:13,810 These cases often involve an interpretation 40 00:02:13,810 --> 00:02:18,110 of the Constitution, and have significant social, political, 41 00:02:18,110 --> 00:02:19,310 and economic consequences. 42 00:02:22,020 --> 00:02:25,190 There have been many significant and groundbreaking decisions 43 00:02:25,190 --> 00:02:27,180 made by the Supreme Court. 44 00:02:27,180 --> 00:02:30,610 These are a few notable decisions that were made. 45 00:02:30,610 --> 00:02:33,500 In 1942, the Supreme Court decided 46 00:02:33,500 --> 00:02:36,250 on the Wickard v. Filburn case. 47 00:02:36,250 --> 00:02:39,360 This case recognized the power of the federal government 48 00:02:39,360 --> 00:02:42,050 to regulate economic activity. 49 00:02:42,050 --> 00:02:44,060 Filburn was a farmer, who was growing 50 00:02:44,060 --> 00:02:46,510 wheat for on-farm consumption. 51 00:02:46,510 --> 00:02:50,000 However, the US had established limits on wheat production, 52 00:02:50,000 --> 00:02:53,420 and Filburn was exceeding those limits. 53 00:02:53,420 --> 00:02:55,930 So even though the extra wheat he was producing 54 00:02:55,930 --> 00:02:59,350 was for his own use and he had no intention of selling it, 55 00:02:59,350 --> 00:03:02,110 he was forced to destroy it. 56 00:03:02,110 --> 00:03:06,380 In 1973, the Supreme Court decided on the Roe v. Wade 57 00:03:06,380 --> 00:03:10,470 case, one of the most well-known cases to this day. 58 00:03:10,470 --> 00:03:13,790 They decided to legalize abortion, and by doing this, 59 00:03:13,790 --> 00:03:16,570 prompted a national debate that continues today 60 00:03:16,570 --> 00:03:19,680 about the legality of abortion. 61 00:03:19,680 --> 00:03:22,140 In 2000, the Supreme Court actually 62 00:03:22,140 --> 00:03:25,510 decided the outcome of the presidential election. 63 00:03:25,510 --> 00:03:28,190 The race was so close in the state of Florida, 64 00:03:28,190 --> 00:03:31,020 that a recount of the ballots was required. 65 00:03:31,020 --> 00:03:33,610 But the Florida Secretary of State certified 66 00:03:33,610 --> 00:03:36,600 that President Bush was the winner before the recount 67 00:03:36,600 --> 00:03:38,190 could be completed. 68 00:03:38,190 --> 00:03:40,470 The case then went to the Supreme Court 69 00:03:40,470 --> 00:03:44,030 where it was ruled that all ballots needed to be recounted. 70 00:03:44,030 --> 00:03:46,329 But since this could not be done before the winner had 71 00:03:46,329 --> 00:03:49,500 to be declared, President Bush won the state of Florida, 72 00:03:49,500 --> 00:03:52,210 and thus, the presidency. 73 00:03:52,210 --> 00:03:55,550 A very recent case from 2012 dealt with the Patient 74 00:03:55,550 --> 00:03:58,390 Protection and Affordable Care Act, commonly 75 00:03:58,390 --> 00:04:01,580 called ObamaCare, which requires most Americans 76 00:04:01,580 --> 00:04:03,160 to have health insurance. 77 00:04:03,160 --> 00:04:05,280 The Supreme Court upheld this requirement. 78 00:04:08,220 --> 00:04:10,860 Since non-profits, voters, and anybody 79 00:04:10,860 --> 00:04:12,830 interested in long-term planning can 80 00:04:12,830 --> 00:04:16,089 benefit from knowing the outcomes of the Supreme Court 81 00:04:16,089 --> 00:04:19,560 cases before they happen, legal academics 82 00:04:19,560 --> 00:04:22,010 and political scientists regularly 83 00:04:22,010 --> 00:04:24,750 make predictions of Supreme Court decisions 84 00:04:24,750 --> 00:04:29,260 from detailed studies of the cases and individual justices. 85 00:04:29,260 --> 00:04:32,090 In 2002, Andrew Martin, a professor 86 00:04:32,090 --> 00:04:34,940 of political science at Washington University in St. 87 00:04:34,940 --> 00:04:37,520 Louis, decided to instead predict 88 00:04:37,520 --> 00:04:41,780 decisions using a statistical model built from data. 89 00:04:41,780 --> 00:04:44,040 Together with his colleagues, he decided 90 00:04:44,040 --> 00:04:47,170 to test the model against a panel of experts. 91 00:04:47,170 --> 00:04:49,550 They wanted to see if an analytical model could 92 00:04:49,550 --> 00:04:52,690 outperform the expertise and intuition 93 00:04:52,690 --> 00:04:53,970 of a large group of experts. 94 00:04:57,030 --> 00:05:00,150 Martin used a method called classification and regression 95 00:05:00,150 --> 00:05:02,250 trees, or CART. 96 00:05:02,250 --> 00:05:04,620 In this case, the outcome is binary. 97 00:05:04,620 --> 00:05:08,650 Will the Supreme Court affirm the case or reject the case? 98 00:05:08,650 --> 00:05:11,720 He could have used logistic regression for this, 99 00:05:11,720 --> 00:05:15,900 but logistic regression models are not easily interpretable. 100 00:05:15,900 --> 00:05:18,780 The model coefficients in logistic regression 101 00:05:18,780 --> 00:05:22,600 indicate the importance and relative effect variables, 102 00:05:22,600 --> 00:05:24,900 but do not give a simple explanation 103 00:05:24,900 --> 00:05:27,510 of how a decision is made. 104 00:05:27,510 --> 00:05:30,660 In this lecture, we'll discuss the method of CART, 105 00:05:30,660 --> 00:05:33,480 and a related method called random forests. 106 00:05:33,480 --> 00:05:36,120 We will then see if these methods can actually 107 00:05:36,120 --> 00:05:40,380 outperform experts in predicting the outcome of Supreme Court 108 00:05:40,380 --> 00:05:41,790 cases.