Economics 698 Information Markets and the Wisdom of …

[Pages:17]Economics 698

Information Markets and the

Wisdom of Crowds

Notes

David L. Kelly

Department of Economics University of Miami Box 248126

Coral Gables, FL 33134 dkelly@miami.edu

Current Version: Fall 2018

INTRODUCTION: INFORMATION MARKETS AND THE WISDOM OF CROWDS

I Idea for the Course

A Course Description Markets are powerful institutions for organizing buyers and sellers, aggregating information, and managing risk. Many companies are now creating their own markets specifically tailored to obtain particular information (for example, will a construction project be completed on time?) or assess a particular risk (what is the probability of a foodborne illness at a restaurant chain?). Such markets rely on the "wisdom of crowds":

Definition 1 Wisdom of Crowds: A theory that a large number of individuals (a crowd), each with limited information, forecasts better than an expert, with more information than any single individual.

The course will analyze different types of information markets, including prediction markets, polls and surveys, auctions, and existing securities markets. Students will analyze strengths and weaknesses of each approach relative to traditional expert opinion. Students will consider applications of information markets, such as risk management, crowdfunding, and mergers.

B Limited Information Most managerial decisions are made with very limited amounts of information. Examples:

1. Market Entry: Firms decide whether or not to enter a new market based on limited information. The firm may do a limited survey of consumer preferences, but has no actual experience in the market to guide decisions. In addition, information such as how incumbent firms will react (cut their prices, etc.), whether other firms will also enter, and potential production and distribution difficulties are difficult to estimate.

2. Hiring: Mangers make hiring decisions often based only on a one page resume and a few hours of interviews. The actual quality of the employee is uncertain.

3. Bank Lending: how much does a bank really know about borrowers?

1

4. Inventory Decisions: When ordering inventory, how much does the firm know about customer demand? The firm is uncertain about what styles will consumers buy, what colors consumers prefer, and what sizes.

5. Investment Decisions: An investor buying an S&P 500 index fund has only limited information as to what the return on the investment will be.

It is helpful to characterize in what ways information is limited.

Definition 2 unknown knowns: risks that have been correctly identified and properly measured.

? The risk is quantifiable in the sense that the risk can be measured using past data.

? Miami International Airport (MIA) and American Airlines might use data on past hurricanes to estimate the probability of a storm large enough to force the closure of MIA.

? The volatility of investment returns of a particular index (e.g. S&P 500) can be measured using past data.

? In the bank lending case, using past data, banks estimate credit risk based on attributes of borrowers, such as credit history.

? From previous seasons, a clothing retailer knows approximately what percentage of t-shirt sales are small, medium, and large.

Definition 3 Known unknowns: risk that have not been measured or are not easy to measure, but are known to exist.

? The risk is not easy to quantify, because it is rare or because data is lacking.

? American Airlines might know that future storms might be more frequent/severe than past storms, perhaps due to climate change. Past data does not account for this, but American knows that more severe storms are possible.

? Most estimates of the S&P 500 volatility use statistical methods that effectively rule out "black swan" events which may cause a large negative return (e.g. a mortgage crises). Investors (should) know that black swan events are possible and that they are not accounted for in standard volatility measures.

2

? The bank knows that in a severe recession, even borrowers with good credit histories may default.

Definition 4 Unknown unknowns: risk that have not been measured, and are not known to exist.

? Managers and firms may be completely unaware that, after entering a market in a foreign country, that a new government may arise which is hostile to foreign investment.

? American Airlines may be completely unaware of the possibility that the airport could be shut down due to a computer failure, terrorism event, or natural disaster.

When Delta Airlines had a computer failure in 2016 which grounded flights for several days, it lost $150M. A power surge cut off power to the computer system, but the backup computer system did not turn on. If Delta had been aware of that their backup system did not work, they would surely have fixed the problem, which would have cost far less than $150M. So managers making key decisions were probably unaware of the potential problem.

Or consider in 1991, when a farmer burying a cow cut through a fiber optic cable knocking out 20 air traffic control centers. Most likely, this was an unknown unknown.

The idea for this course: How can managers acquire more information, to understand better known risks and to learn about the existence of unknown risks?

Answer: Rely on the wisdom of crowds. It is certainly possible that a large number of people, each with a little information, can add up to more information than an expert has. But how can a manager find a crowd and how can the manager extract the information that the crowd has? That is, how can the information be aggregated?

II Sources of Information

A Past Data

It is common for managers to use past data to reduce uncertainty. For example, American Airlines can use past data on storms in Miami to estimate the probability of a hurricane closing Miami International Airport. Consider:

? Number of direct hits to Southeast FL of category 1 or higher hurricanes category (1851-2017): 49.

3

? Years studied: 2017-1851 = 166.

? Average per year: 49/166 = 0.295.

? Interpretation: the probability of a category one or higher hurricane making a direct hit in Southeast FL in a given year is 29.5%.

Using past data has advantages and disadvantages.

? Minimum uncertainty: under certain conditions, the sample mean makes the best use of the data in that it reduces uncertainty as much as possible. This will be studied in more detail later.

? Unbiased prediction: under certain conditions, the prediction of 29.5% will not be consistently too high or too low.

? Unknowns are not accounted for. For example, we might know that climate change will mean the probability of a hurricane in the future is different than the past. The past data then becomes biased too low.

? Information is limited. For example, physical models based on water temperature, wind patters, and weather events such as El Nino may predict better than past data (this is certainly true for predicting landfalls once a hurricane appears). The estimate of 29.5% uses only past data, not data from physics models.

B Expert Opinion

Another common strategy is to go with the opinion of managers with expertise in the uncertainty. For example, a manager of a clothing store may have years of experience and so have a pretty good idea as to what sizes are the most common. In the movie Moneyball, the team initially relied on scouts with years of experience to ascertain which prospects were most likely to succeed. The firm can bring in a recruiting consultant to go through resumes and find the best candidate.

? Expert Experience. Similar to past data, expert has experience which can lead to good predictions.

? Can be biased. The expert may have incentives to give biased reports. For example, if the expert receives a bigger budget when claiming sales will be great, we can expect overly optimistic predictions.

4

? Many unknowns are not accounted for. The expert is only one person, and cannot know everything.

? Many sources of information are not accounted for. The expert may not have seen some information.

? Herd mentality. Sometimes experts do not want to predict differently than all the other experts. Being wrong in this case can cause a loss of reputation as an expert.

It is easy to find examples where the experts get it wrong. Of course, it is OK to be wrong. No one can have perfect information. The question is whether or not the expert is biased and whether or not other sources of information are more accurate.

C Surveys and Polls

One can conduct a poll or survey to gain information. In this class, we will use the terms poll and survey interchangeably. One might conduct a survey to see if consumers will like a new product, for example. Surveys come in many forms, including small focus groups.

? Diverse Opinions. By surveying a large group, the manager gets diverse opinions. Less susceptible to the herd mentality.

? Some unknowns are accounted for. The manager can learn things he/she does not know, but those being polled do.

? Opinions are not weighted. Some survey respondents may have strong opinions and others may not care, but each respondent receives only one vote.

? Those being surveyed lack incentives. Since there is no monetary reward for getting a correct answer, respondents have a tendency to give answers without much thought. For this reason, surveys are not very useful for questions like "what will the price of Google stock be next year?". However, certain questions regarding, for example, taste and preferences do not require much thought. For example, do you like the taste of yogurt?

? Sampling is critical. Results will be biased if, for example, only people with a lot of time on their hands answer surveys. Biases can occur based on how the question is phrased, based on the interviewer, and based on who responds.

5

D Securities Markets

An efficient market has a very desirable feature. In particular, at any point in time market prices of securities provide accurate signals for resource allocation.

Fama and Miller (1972)

if the capital market is to function smoothly in allocating resources, prices of

securities must be good indicators of value.

Fama (1976)

Stock markets and other securities markets convey information. When the price of a stock increases, this signals that traders in the market have increased their beliefs about future cash flows. Indeed, the revelation of information is one of the most important aspects of a securities market. When the price of the stock increases, this signals information to the firm that it is on the right track. It signals to investors that the firm is a good place to invest in. It even signals to potential entrants that the firm is producing a product that consumers want. Investment dollars are naturally directed to the most productive firms.

Consider an example, a recent announcement that T-Mobile and Sprint will merge, on April 29, 2018. When the markets opened on Monday, April 30, both stock prices fell. TMobile stock fell by 6.2% and Sprint stock fell by 13.7%. The collective wisdom of all the traders in the market, presumably none of whom have access to the information that Sprint and T-Mobile executives have, was that the value of T-Mobile's cash flows would fall because of the merger.

6

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download