Chapter 4, Problem 1.2VC

For its first 2 decades of existence, the NBA’s Orlando Magic basketball team set seat prices for its 41-game home schedule the same for each game. If a lower-deck seat sold for $150, that was the price charged, regardless of the opponent, day of the week, or time of the season. If an upper-deck seat sold for $10 in the first game of the year, it likewise sold for $10 for every game.

But when Anthony Perez, director of business strategy, finished his MBA at the University of Florida, he developed a valuable database of ticket sales. Analysis of the data led him to build a forecasting model he hoped would increase ticket revenue. Perez hypothesized that selling a ticket for similar seats should differ based on demand.

Studying individual sales of Magic tickets on the open Stub Hub marketplace during the prior season, Perez determined the additional potential sales revenue the Magic could have made had they charged prices the fans had proven they were willing to pay on Stub Hub. This became his dependent variable, y, in a multiple-regression model.

He also found that three variables would help him build the “true market” seat price for every game. With his model, it was possible that the same seat in the arena would have as many as seven different prices created at season onset—sometimes higher than expected on average and sometimes lower.

The major factors he found to be statistically significant in determining how high the demand for a game ticket, and hence, its price, would be were:

► The day of the week (x₁)

► A rating of how popular the opponent was (x₂)

► The time of the year (x₃)

For the day of the week, Perez found that Mondays were the least-favored game days (and he assigned them a value of 1). The rest of the weekdays increased in popularity, up to a Saturday game, which he rated a 6. Sundays and Fridays received 5 ratings, and holidays a 3 (refer to the footnote in Table 4.2).

Fernado Medina

TABLE 4.2 Data for Last Year’s Magic Ticket Sales Pricing Model

His ratings of opponents, done just before the start of the season, were subjective and range from a low of 0 to a high of 8. A very high-rated team in that particular season may have had one or more superstars on its roster, or have won the NBA finals the prior season, making it a popular fan draw.

Finally, Perez believed that the NBA season could be divided into four periods in popularity:

► Early games (which he assigned 0 scores)

► Games during the Christmas season (assigned a 3)

► Games until the All-Star break (given a 2)

► Games leading into the play-offs (scored with a 3)

The first year Perez built his multiple-regression model, the dependent variable y, which was a “potential premium revenue score,” yielded an R² =.86 with this equation:

$y=14,996+10,801x_1+23,397x_2+10,784x_3$

Table 4.2 illustrates, for brevity in this case study, a sample of 12 games that year (out of the total 41 home game regular season), including the potential extra revenue per game (y) to be expected using the variable pricing model.

A leader in NBA variable pricing, the Orlando Magic have learned that regression analysis is indeed a profitable forecasting tool.

2. Use the data to build a model with rating of the opponent as the sole independent variable.