Media Bias: Direction and Variance

Michael Seringhaus touches on an interesting point regarding media bias. In terms of politics, whatever long-term directional biases the media has, in the short run its "bias" is also towards more variance.

It should be clear to everyone that what is reported during campaigns is often fed to the media by the camps themselves. This doesn't just apply to facts and trial-balloons, but also arguments that are more editorial in nature. For one thing, this dynamic allows reporters to ingratiate themselves to the campaigns, who can rely on them periodically to push their respective talking points. In exchange, reporters get the occasional tip. The media can act like a double-agent in this respect, alternating directional bias from news-cycle to news-cycle, week to week.

Most importantly, this "variance" makes the contest seem closer and more exciting than it may be in reality, which encourages consumption of the news product. If you were watching the markets, this tendency manifested itself most obviously on the election night. Long after Intrade gave a 99% chance for an Obama victory, CNN announced him as the winner, predictably at 11pm EST on-the-dot.

Furthermore, it turns out that prediction market prices near 50% might inherently be more volatile. This is the argument of a new paper by Panos Ipeirotis and Nikolay Archak in which they introduce a model where prediction market volatility depends only on price and time to expiration, as summarized in the below chart. As a trader, I think I would rely more on historical volatility in most cases for a couple of reasons, but the model is intuitive and suggestive with respect to the "news product" idea.

Tax Futures: Cutting to the Chase

For most people the election outcome will manifest itself most directly in the form of taxes. We can easily ascertain the general platforms of each candidate on this issue, but the specifics are intentionally left a little vague during the campaign season. Since they launched in February, the Intrade tax contracts help to cut through candidates' omissions, and provide a direct hedge to this pervasive risk. It's good to know who will win the election, but it's better to know how the result will affect you -- and be able to do something about it.

The tax contracts can be interpreted as revealing the probabilities of tax changes conditional on who wins the election. For example, currently they give a 61% chance of the top marginal rate for single filers going up in 2009. If we assume this would be impossible under McCain's veto power, since Obama's chances for the presidency are 84%, this implies a 73% percent chance of taxes being raised in 2009 if Obama wins. That is, P(tax hike GIVEN Obama win) = P(take hike AND Obama win) / P(Obama win). Once the election is over, the contracts will continue to reflect the likelihood of legislation passing and other fiscal developments that might spur policy.

Of course, these specific markets are still young enough that marginal buyers and sellers can have a significant influence on prices. Care must be taken with such interpretations, but these issues should be greatly mitigated in future instantiations once US regulations are modernized.

Contractability will always be a challenge though, that is, making sure the hedge actually hedges. We really only care about our effective rates, and the highest marginal rate is only part of that equation. The many specific credits in Obama's tax plan underline this challenge. (I thought his advisor Austan Goolsbee was a proponent of tax code simplicity?) Still, while there is basis risk in the contract design, the highest marginal rate is the most tractable proxy for effective rates, especially for who one imagines to be the typical Intrade investor.

Going forward, it's imperative that policy-makers do not view such markets with hostility. It's difficult enough to design a contract that hedges an event outside of anyone's control. We don't want to get into a situation where legislators are reacting to the design of contracts and frustrating their usefulness with unexpected changes. (Clearly contracts should refer to outcomes like tax rates, and not specific bills.) We have to continue to demonstrate the desirability of these markets both in terms of price discovery and hedging. Policy-makers should consider them to be helpful insofar as they: (1) are informative about the consequences of competing policies, as with Robin Hanson's "decision markets", and (2) ease log-jams through their hedging function that allows private interests to essentially meet each other half way.

To this end, it is encouraging to see Austan Goolsbee write things like this on the advantages of market predictions. It is even more encouraging to hear from Jason Furman, Obama's Economic Policy Director, that the most important question to Obama is, "What does Paul Volcker think?" But why do I get the feeling that is hyperbole offered to placate certain undecideds?

The Benefits of Market Forecasts

At this point in the election lead-up, many people probably find themselves focused on the differences between the various projections out there. While the forces of supply and demand sometimes influence Intrade prices in unexpected ways, markets have several advantages over other means of forecasting. None of what I'm going to say is news, but it bears repeating, especially in this environment where people are suspicious of market mechanisms.

First, non-market methods of predicting are vulnerable to over-optimization. If someone fiddles with a model and its weights, they are likely to produce something with an excellent track record in retrospect, but what worked best in the past might not work in the future. Markets can be influenced by well-funded traders, but are less vulnerable to the errors in judgment of isolated individuals.

Second, markets naturally take non-market projections as inputs. If a non-market methodology works well, markets will gradually incorporate its output. While the opposite is possible in theory, someone would need to decide what weights to assign the market and non-market components, which brings us back to the first point. Markets have the advantage of flexibly incorporating all available forecasts.

Many have noted that lack of transparency in asset pricing (not to mention models) played a large role in the credit crisis. Exchange-traded markets provide real-time price transparency, 24/7 in the case of Intrade. Historical price and volume data allow for additional analysis, including detection of when traders may be exercising market power. Non-market methods are likely to only provide a snapshot of their output, making systematic analysis by users
difficult. This may also bias reports of success. If a non-market analysis beats the market, you can be sure you'll hear about it. The opposite is unlikely.

It may be that in the case of elections, poll-based models give markets a run for their money, but that is a specific domain with a relatively rich dataset. It bears noting again that markets provide incentives for uncovering new information, an activity that is quickly reaching maturation when it comes to elections.

Finally, you may not even care about trying to forecast a result you can't control, but markets allow you to control how you are affected by the result. Modernization of regulations in the US will lead to far greater liquidity for prediction markets, which will finally allow for meaningful political hedging. Ultimately, risk-sharing markets may even help to address issues with representative democracies such as the "special interest" problem.

That last bit may sound crazy as I write this in October 2008, but look, ultimately what we are talking about are proper incentives and information aggregation, the lack of which contributed massively to the failure of 20th century communism. If 20th century finance institutions became lacking in the same areas, do you fault "markets" in the abstract?