What I will do is the following and I would appreciate your feedback on my approach or if you suggest another approach.
1- Start with the whole population
2-segment based on age (children, middle age, old people).
3- Assume all children eat a candy a day (is it underestimated?) , then multiply by number of days in a year.
4-Assume middle age only 30% would eat candy and they only eat one candy per week, then multiply time number of weeks in a year.
5- Assume old people won't eat candy.
6- Add number of candies eaten by children and middle aged to get total candy consumption.
please give your feedback on approach and feel free to suggest any other approach.