Pfizer: Betting on Success

The Ivey Business Review is a student publication conceived, designed and managed by Honors Business Administration students at the Ivey Business School.

A Plethora of Pharmaceuticals

Pfizer was founded in 1849 by cousins Charles Pfizer and Charles Erhart. In the 170 years since, the company has commercialized and distributed medical breakthroughs, growing into a pharmaceutical giant. Pfizer offers more than 370 products ranging from recognizable brands like Advil and Robitussin to more exotic concoctions like antiemetics for chemotherapy and antibiotics. In 2018, this product mix generated revenues of $53.6 billion for the company.

Pfizer’s success is largely attributable to its investment into the research and development of new drugs. The company is a leader in drug development and had 10 FDA approvals throughout 2017. While a key factor contributing to Pfizer’s market dominance, drug development carries inherent risk and there is no guarantee that any experimental drug will make it to market. Pfizer’s willingness to carry this risk, as evidenced by its continuous increase in research and development (R&D) investment, serves to further accelerate the development and delivery of potential blockbuster treatments to individuals around the world.

Volatility in R&D Expenses

In 2013, researchers analyzed a series of new drugs brought to market between 2002 and 2011, comparing analyst forecasts and real-world results. The study revealed that sales forecasts for these new pharmaceutical products were wildly inaccurate, with 60 per cent of analyst consensus forecasts having a margin of error greater than 40 per cent. Furthermore, a study by Citi Research concluded that two-thirds of novel drugs that make it to market fail to meet analyst expectations for first-year revenues and that a strong correlation exists between first-year sales and future performance. Given that pharmaceutical companies allocate R&D budgets based on forecasted drug sales, poor forecasting can lead to substantial losses.

The issue of unreliable forecasting must be addressed soon. Industry R&D expense growth has exceeded revenue growth for the past decade, indicating a need for stronger budget allocations. Pfizer’s R&D expenses were $8.01 billion for fiscal 2018 and have increased by four per cent since 2017, outpacing revenue growth of two per cent over the same period.

To improve its forecasting accuracy, Pfizer should implement a prediction market to be used by internal analysts, salespeople, and other employees. Prediction markets have been successfully used by other companies to improve the effectiveness of management decisions. Such technology holds promise to be used for an analogous purpose in this novel market.

What is a Prediction Market?

A prediction market is a platform where individuals can trade contracts with payoffs linked to the binary outcome of a specific event. For instance, a contract could be created with payoff tied to the outcome of a fair coin toss: if the coin lands heads up, the holder receives $100. If it lands tails up, the holder receives nothing. A 50-per-cent chance of receiving $100 translates to an expected value of $50, so one would expect the contract to trade at a price close to $50. This simple example illustrates a basic principle of prediction markets: the amount that traders are willing to pay for a given contract corresponds to their subjective estimate of an event’s likelihood of occurring.

Contracts can be assigned payoffs based on any binary outcome imaginable. The premise behind prediction markets is the aggregation of information and beliefs; encouraging employees from different departments with access to different information to trade contracts allows the market to converge on the most accurate predicted value of the contract. Oftentimes, these markets allow companies to factor in information that employees may otherwise avoid discussing—for example, project delays.

Prediction markets are currently being used by companies like Google and Intel to forecast new product sales and even the likely outcome of new research and innovation activities. Overall, the use of these prediction markets has yielded positive results: Hewlett-Packard claims that BRAIN, the internal prediction market it piloted for Swisscom, forecasted results 27-per-cent closer to actual financial metrics than polling groups and 17-percent closer than top experts, and Intel reportedly saw its internal prediction market outperform official demand forecasting by 20 per cent.

In the previously referenced study, while 60 per cent of analyst consensus forecasts had an error margin of more than 40 per cent, the median consensus prediction for all drugs studied had only an error of four per cent. This indicates that even if the high variance of forecasts makes any one analyst’s prediction unreliable, aggregating results yields a prediction close to reality. Prediction markets are uniquely suited to this situation since they aggregate a wide array of perspectives and encourage precise quantification of subjective belief. Furthermore, the historical performance of prediction markets at forecasting other metrics such as the success of new sales initiatives suggests that they would also be able to predict sales of new drugs with a relatively high degree of accuracy.

Pfizer’s Prediction Market

The new prediction market can be used by Pfizer analysts and salespeople to predict the success and sales of drugs in the development pipeline. Participants will use imitation currency to purchase contracts from one another based on their forecasts for expected market success of a new drug. Google, for example, used “Goobles” as currency in its prediction markets and rewarded small prizes to employees that accumulated the highest amounts of Goobles. Similarly, top forecasters in Pfizer’s prediction markets could earn small prizes as their forecasts are proven correct, providing a pecuniary incentive to trade based off of one’s best inclination.

While one might suspect that trading with imitation currency would lead to less accurate information, empirical data shows that this is not so. A 2008 study in The Journal of Prediction Markets found that valid and robust results were obtained whether real or imitation currency was used.

Pfizer could ask its prediction market: “Will sales of Drug X exceed $1 billion in 2019?” Using imitation PfizerBucks (PBs), participants could then purchase two contracts, with one giving the holder a 100 PB payout if Drug X exceeds $1 billion in sales, and the other giving the holder a 100 PB payout if sales fall below $1 billion. Participants could buy and sell either contract on the internal market, with those who believe in Drug X trying to accumulate contracts paying out if it succeeds and vice versa.

Pfizer would look at the price of each contract to see what the market in aggregate believes Drug X’s odds of success to be. If the market price is 80 PB for this contract, then the market believes there is an approximate 80-per-cent chance that sales will indeed exceed $1 billion. Pfizer can use that information to decide whether or not to launch the drug. In the case where market prices indicate a very low probability of success and management decides not to proceed with an initiative, the contracts will pay out as if the event occurred, but was unsuccessful.

The new prediction market strategy should be implemented parallel to Pfizer’s existing forecasting processes. Additionally, Pfizer should collect employee feedback and work with the selected vendor to make continuous improvements to the platform.

Concerns About Prediction Markets

Cunning employees could trade contracts in order to manipulate market prices and influence the outcome to their benefit. For example, an employee in research could stock up on contracts for one of his or her clinical tests to fail, and then skew the results so that the contract will pay out. To prevent such manipulation, limits should be established on individual positions, limiting the benefit that could be realized and consequently reducing the incentive to engage in dishonourable behavior. Individuals with external compensation directly tied to the outcome of the event should also be prevented from participating in the market for similar reasons.

Despite the fact that some information may be limited by the exclusion of certain participants, prediction markets have a track record of maintaining a high level of accuracy thanks to a “wisdom of crowds” effect. Most notably, a famous experiment conducted by researcher Philip Tetlock demonstrated that the aggregation of predictions from laypeople spread throughout the U.S. could outperform intelligence analysts with access to classified information in the forecasting of complex geopolitical events.

Another risk associated with this strategy is the need for many participants to leverage the “wisdom of crowds” effect and ensure prediction accuracy. Public prediction markets with well-calibrated forecasting in the realm of computers and internet, politics, and science have been shown to have as low as 27 market participants without compromising accuracy. With more than 90,000
employees, it is unlikely that Pfizer would find itself with a shortage of participants.

Additionally, prediction markets face legal issues given their inherent similarities to gambling. Avoiding the use of real currency in the company’s prediction market would mitigate this risk, as it parallels the implementation plan of other companies’ prediction markets. Contracts traded on internal company markets do not qualify as securities and thus would not be regulated by the U.S. Securities and Exchange Commission.

Forecasted Fortunes

Sales forecasting is but one application of prediction markets in the pharmaceutical industry. Economists such as Robin Hanson, one of the earliest proponents of prediction markets, are exploring the applicability of these markets for forecasting the outcome of scientific study replications. This could provide value to pharmaceutical companies when integrated into their development pipelines. Other applications may include identifying promising new drug solutions and compounds worth investigating, forecasting participant enrolment in laterstage clinical trials, and forecasting the likelihood of new regulations that could impact Pfizer’s business. If Pfizer is able to harness the “wisdom of the crowds,” the pharmaceutical giant will be one step closer to overcoming one of the pharmaceutical industry’s greatest problems, the volatility of R&D risk.