Sports attendance and rain

Second-year geography students at the Australian National University in 1996 examined the effect of rain on attendances at outdoor sporting events. Five of the reports are outlined below.

1. Greg Palethorpe used data for the main Canberra rugby-league team playing 55 matches at home during a different period, 1990-1994. His data showed a mean attendance of 14,442 on the 40 days without rain, 15,053 on the 7 days when rain was no more than 1.2 mm, 12,384 on the 4 days with between 1.2 and 6 mm, 12,667 on 4 days with over 6 mm. These figures show an ill-defined effect of rain on attendance. Incidentally, it is important to use rainfall measurements taken at 9am on the day AFTER the match, to find rainfall during the period of the match itself.
2. Jeremy Howes examined the attendances at 33 home matches of the same team during 1993-1996. In this case, the attendance figures were normalised twice to allow for i) the position of the game in the year's series of 11 games (the biggest attendance is for the final game, for instance), and ii) the year-to-year decline in popularity of the game (1). On the 20 days without rain, the normalised attendance was 110% of the mean, on the 5 days with not more than 1 mm of rain the figure was 89%, on the 6 days with 1 - 6 mm 98%, and on the 2 days with 6 - 12mm 63%. In other words, normalisation shows more vividly the effect of rain on attendance numbers.
3. A more sophisticated procedure was adopted by Christine Nicholas in analysing data on attendances at 42 Australian Football matches at the Melbourne Cricket Ground, involving either the North Melbourne or the Melbourne clubs. It was hypothesised that attendance depended on the four factors - rainfall, the average of respective popularity ratings of the two teams in the match, whether a Melbourne team is involved, and whether or not the match is the final in a competition. The popularity rating of each team was 1, 2 or 3, according to the team's position on the competition ladder and the size of the club. An additional unit was added for a Melbourne team and one more if the game was a final. Attendances ranged between 12,783 (for a match between Melbourne and Brisbane) and 93,102 (for a final between North Melbourne and Sydney). It rained on 22 of the 42 days. Then multi-factor regression analysis was applied to the relationship between attendance figures and values for the four variables. This demonstrated the importance of the factors other than rainfall. Rainfall did reduce attendance, but hardly significantly.
4. The double normalisation technique was applied by Alison Hamilton to data on attendance at the Melbourne Spring Horse-racing Carnivals between 1985 - 1995. Each Carnival takes place over four days, the second being the day of the famous Melbourne Cup race. The data were first normalised by the day of the week (i.e. each of the raw numbers was divided by the average of attendances on that day for all Carnivals). This removed the effect of the large attendances on the second day. Then values for any day were divided by the average for all four days in that year's Carnival, to remove year-to-year differences. The result was a demonstration that in this case there was no clear effect of rain amount on the day of racing (until 9am the following day) on the attendance. It appears too that the amount of rainfall in the previous 24 hours (ending at 9am on the race day) had no consistent effect.

(1) Linacre, E.T. 1990. The effect of rain on attendance at Sydney's Easter Show. Aust. Meteor. Mag., 38, 65-7.