Font Size: AAA // Print // Bookmark

Comment for Industry Filing 22-002

  • From: David Rothschild
    Microsoft Research

    Comment No: 69735
    Date: 9/22/2022

    Comment Text:

    We are academic researchers who study prediction markets for both the value they provide in understanding the real-world events that they predict, and what they teach us about market design and usage that is widely applicable to numerous fields. We are writing in favor of allowing Kalshi (or any similar entity) to offer a broad range of political and policy event futures, including the election outcomes they are currently proposing.

    Prediction markets work because they ask the right questions of the right people, who are properly incentivized both to answer them honestly and come back and update their positions when new information becomes available to them. Statistical models work very well in situations where there is high repetition along with stable and available data (for example: frequently companies can predict daily sales numbers in stable industries very well from historical sales data), but are untenable if the outcome or necessary data is idiosyncratic (for example: predicting the sales for one day at random pop-up stores or creating predictions of sales when the sales data is captured differently by store). Polling works very well at getting a snapshot of the people available to answer a given poll, but it is not a prediction of what will happen in a larger target population (for example: a poll cannot take into account unreachable populations or expected changes between now and the outcome of the event). Further, while polling was relatively stable for decades from the 1950’s to 1990’s, dramatic shifts in how pollsters try to reach people due to shifting technology, lower response rates, and increasing correlation with non-response and outcomes of interest have raised additional concerns about the quality and consistency of polling in recent years. Prediction markets take advantage of both models and input data like polling, but they also motivate experts to aggregate that available information along with dispersed information, and intuition about how idiosyncratic information will affect outcomes as the events unfold. And, by aggregating many independent experts together, using their marginal willingness to pay to help weigh them, prediction markets do a great job in making predictions in idiosyncratic situations, such as found in political and policy events.

    Authors of this letter have written extensively in the academic and popular press about prediction markets.* We have documented how prediction markets-based predictions outperform other key predictions in: accuracy, latency, and time-granularity.** As a result, market-based predictions are uniquely impactful in event studies, such as politics and policy. Further, prediction markets are nimble and transparent, culminating with a pricing event, making them particularly attractive for research on how market design affects trading on various conditions.*** These learnings help improve the efficiency of a wide range of markets.

    Prediction market prices in political and policy events would help facilitate price discovery in a wide-range of asset markets, affecting the entire economy (note that pricing is freely available to non-traders). Political and policy events matter: they expose a wide-variety of businesses to risk that traditional financial markets have trouble pricing. A robust set of markets for political and policy events could price that risk, and, if they were allowed to flourish, could eventually grow to provide hedges where uncertainty is particularly acute.

    Concerns that these types of markets could cause manipulations in the outcome, or be manipulated, are misplaced. First, the market caps are many magnitudes smaller than the amount of money influenced by these political and policy events: stakeholders with the ability to affect events will not be incentivized by the relatively small amount of money they could make investing against their public interests. Second, manipulating prediction market prices has proved to be very hard, transparent, and relatively short lived.**** With a transparent order book it is very easy to see if someone is attempting to manipulate a market, immediately mitigating the impact of any short-lived price manipulation. Thus, manipulations have had little impact on the derived underlying probability of the event, by those who follow the prices.

    Harry Crane, Professor, Department of Statistics, Rutgers University
    David M. Pennock, Director, DIMACS Center, and Professor, Department of Computer Science, Rutgers University
    David Rothschild, Economist, Microsoft Research, and Fellow, CSS Lab, University of Pennsylvania
    Koleman Strumpf, Burchfield Presidential Chair of Political Economy, Department of Economics,Wake Forest University

    *Authors of this letter are author(s) on all of the papers referenced, which represent a small percentage of their body of work on the prediction markets.
    **See research examples: Rothschild (2009), Rothschild (2015), Crane (2019), Crane and Vinson (2022), Strumpf and Rhode (2004).
    ***See research examples: Rothschild and Pennock (2014), Rothschild and Sethi (2016), Schmitz and Rothschild (2019)
    **** See research example: Strumpf and Rhode (2008)

    Chen, Yiling, Stanko Dimitrov, Rahul Sami, Daniel M. Reeves, David M. Pennock, Robin D.
    Hanson, Lance Fortnow and Rica Gonen. "Gaming prediction markets: Equilibrium strategies with a market maker." Algorithmica 58.4 (2010): 930-969.

    Crane, Harry and Darrion Vinson. "Models vs. Markets: Forecasting the 2020 U.S. election." Researchers.One (2022),

    Crane, Harry. Polls, Pundits, or Prediction Markets: An assessment of election forecasting. Researchers.One (2019),

    Dudík, Miroslav, Sebastien Lahaie, David M. Pennock, and David Rothschild. "A combinatorial prediction market for the US elections." Proceedings of the Fourteenth ACM Conference on Electronic Commerce, 2013.

    Lambert, Nicolas S., John Langford, Jennifer Wortman Vaughan, Yiling Chen, Daniel M. Reeves, Yoav Shoham, and David M. Pennock. An axiomatic characterization of wagering mechanisms. Journal of Economic Theory, 156 (2015): 389-416.

    Pathak, Deepak, David Rothschild, and Miroslav Dudik. "A comparison of forecasting methods: Fundamentals, polling, prediction markets, and experts." The Journal of Prediction Markets 9.2 (2015): 1-31.

    Rothschild, David. "Forecasting elections: Comparing prediction markets, polls, and their biases." Public Opinion Quarterly 73.5 (2009): 895-916.

    Rothschild, David M., and Rajiv Sethi. "Trading strategies and market microstructure: Evidence from a prediction market." The Journal of Prediction Markets 10.1 (2016): 1-29.

    Rothschild, David, and David M. Pennock. "The extent of price misalignment in prediction markets." Algorithmic Finance 3.1-2 (2014): 3-20.

    Rothschild, David. "Combining forecasts for elections: Accurate, relevant, and timely." International Journal of Forecasting 31.3 (2015): 952-964.

    Schmitz, James, and David Rothschild. "Understanding market functionality and trading success." Plos one 14.8 (2019): e0219606.

    Servan-Schreiber, Emile, Justin Wolfers, David M. Pennock, and Brian Galebach. "Prediction markets: Does money matter?." Electronic Markets 14.3 (2004): 243-251.

    Strumpf, Koleman, and Paul Rhode."Historical Presidential Betting Markets," Journal of
    Economics Perspectives 18 (2004), 127-142.

    Strumpf, Koleman, and Paul Rhode. "Manipulating Political Stock Markets: A Field Experiment
    and a Century of Observational Data," 2008. mimeo available at