Prediction Market Results. Hye Young You

Size: px

Start display at page:

Download "Prediction Market Results. Hye Young You"

Marjory Haynes
5 years ago
Views:

1 Prediction Market Results Hye Young You

2 A Brief Summary 1 Three Prediction Markets 2 Time Span: Participants: EMR 13 Students + Teaching Staffs

3 Question I Will Obama be re-elected in the 2012 presidential election? EMR 13 Crowd Prediction: 74.43% Intrade.com Crowd Prediction: 72.4% FiveThirtyEight: 90.9% Did an Event Happen? Yes (100%).

4 EMR 13 Prediction Price Chart

5 Intrade Price Chart - Entire Period

6 Intrade Price Chart - Last 40 Days

7 FiveThirtyEight Price Chart

8 EMR13: Investment Amount and Frequency Buy (N=65) Average Investment Investment Sell (N=33) Average Investment

9 EMR13: Investment Patterns Number of Days Before the Election Investment Amount factor(type) Buy Sell

10 Question II Will Obama s popular vote share be higher than 50%? EMR 13 Crowd Prediction: 64.2% Iowa Electronic Market Crowd Vote Share Prediction: 50.5% FiveThirtyEight Vote Share Prediction: 50.8% Did an Event Happen? Yes (100%). Actual Vote Share = 50.5%

11 EMR 13 Prediction Price Chart

12 Iowa Electronic Market Price Chart

13 FiveThirtyEight Price Chart

14 EMR13: Investment Amount and Frequency Buy (N=62) Average Investment Investment Sell (N=36) Average Investment

15 EMR13: Investment Patterns Days Before the Election Day Investment Amount factor(type) Buy Sell

16 Question III Will Brown be re-elected in the 2012 Senate race in Massachusetts? EMR 13 Crowd Prediction: 35.26% Intrade.com Crowd Prediction: 20% FiveThirtyEight: 6% Did an Event Happen? No (0%).

17 EMR 13 Prediction Price Chart

18 Intrade Price Chart - Entire Period

19 Intrade Price Chart - Last 40 Days

20 FiveThirtyEight Price Chart

21 EMR13: Investment Amount and Frequency Buy (N=29) Average Investment Investment Sell (N=47) Average Investment

22 EMR13: Investment Amount and Frequency Number of Days before the Election Investment Amount factor(type) Buy Sell

23 How Prediction Market Works In typical stock markets, a human buyer must be matched with a human seller of stock. For example, if Ken is selling 50 shares at $50/share, there must be a buyer willing to pay that price and vice versa. The price of the stock itself is based on the supply and demand of the finite number of shares in play. These transactions are all handled by computer.

24 How Prediction Market Works Inkling Markets uses an automated market maker to control trading. Inkling does not force a buyer to be matched with a seller, and vice versa. From a trader s perspective, Inkling Markets is always the buyer and seller of shares. Inkling Markets sets the stock price according to demand or lack there of. If a trader buys shares, there is demand for the stock and its price goes up. If a trader sells, there is a lack of demand and the price goes down.

25 How Prediction Market Works

26 How Prediction Market Works

27 Payoffs

28 Payoffs

29 Payoffs

30 Payoffs

31 Payoffs

32 Payoffs - IEM

33 Frequency of Investment 80 Trade Count Participant ID

34 Investment Returns Investment Returns Average Returns

35 Investment Returns Investment Returns factor(type) Loss Profit Ranking in Investment Frequency

36 Obama Winning Chance: Comments. Strong debate performance today Buy Obama shares ($275.99) 10/29. Recent polls how a tighter race than expected Sell Obama shares ($ ) 11/2. This election will surely be within a percentage or two. This is way high for Obama Sell Obama shares ($ )

37 Obama Vote Share: Comments 10/11. Gonna take it home Buy Obama share ($774.74)

38 Brown Winning Chance: Comments 10/29. 37% is an insanely unrealistic prediction. Warren will probably win, but not by that margin, in my opinion anyway Buy Brown share ($341.06) Nate Silver puts Brown s chance at 5% This participant sold $3, amount of Brown shares between But this participant bought $ amount of Brown shares in 11.5 as well.

39 Prediction Markets v.s. Polls If the election were held today, who would you vote fore? Who do you think will win the upcoming election? Both questions strive to aggregate the information that is often widely dispersed among economic actors. Which one is a better method?

40 National Polls

41 Prediction Markets: Intrade

42 State Polls: Pennsylvania

43 Prediction Markets: Pennsylvania

44 State Polls: Ohio

45 Prediction Markets: Ohio

46 State Polls: Florida

47 Prediction Markets: Florida

48 State Polls: Virginia

49 Prediction Markets: Virginia

50 Prediction Markets v.s. Polls On average, prediction markets perform better than polls. Why? Self Selection Incentives Efficient Information Aggregation Larger Samples

51 What went wrong with polls in 2012?

52 What went wrong with polls in 2012? Polls are an indicator of what s going on with voters. Pollsters and organizations make predictions based on polls and other factors. But polls may not be the reflection of reality anymore. There may be a strategic purpose to manipulate the procedures of polling.

53 What went wrong with polls in 2012?

54 What went wrong with polls in 2012? Dick Morris, before the election: We re going to win by a landslide. It will be the biggest surprise in recent American political history. It will rekindle the whole question as to why the media played this race as a nail biter. Dick Morris, after the election: I was dead wrong. BUT... I predicted Romney s landslide to help him win. How do polls affect the election outcome? What factors would be affected by polls?

55 Conclusion Designing a mechanism that could efficiently aggregate dispersed individual preference and information will be the most important thing both in politics and business. Prediction markets often have a number of attractive features: quickly incorporate new information, largely efficient, impervious to manipulation. Sometimes fail due to a lack of traders or trading based on private information. Recent use to test economic models (Snowberg, Wolfers, Zitzewitz, 2012 NBER Working Paper).

Prediction Markets are only Human: Subadditivity in Probability Judgments. Bradley C. Love. University of Texas at Austin.

1 Prediction Markets are only Human: Subadditivity in Probability Judgments Bradley C. Love University of Texas at Austin Abstract Prediction markets establish the probability of future events occurring