Model

Creating a queuing model allows us to predict a wealth of performance measures, including expected mean and maximum times a voter might expect at a given point in the day. Variables in the model can then be altered to observe how performance standards change with added servers, improved service time, or different arrival rates.

By creating a comprehensive model of the voting process, we will mathematically represent three unique events in the voting process: arrival, check-in, voting. Check-in and voting would be treated as two independent, serial services. Each service will be defined by an independent mean service rate – derived from data collected in our study – as well as the number of servers available to facilitate the service.

Finally, distributions around the mean rate of each service will be defined. Using stipulated service rates and distributions, this queuing model will generate information for individual voters throughout a simulated day such as time arrived, time spent in line, and time spent checking in and voting. By running iterative simulations based on a set population and hours of operation, we will calculate frequency distributions of maximum waiting times, waiting time for the day’s last voter, and a distribution of mean waiting times throughout the day.

Sample Outcomes

• An election official in any county could derive a distribution of expected wait times for any given time on an election day. He or she could then explore how changing certain features, like the number of voting booths for instance, is likely to impact wait times.
  
• Coupled with the cost-effectiveness estimates, these “what if” simulations could help policy-makers allocate resources to reduce wait times at the lowest possible cost.

Example of queuing model output (in Edelstein):¹

                       

---

¹   Edelstein, W. “New Voting Systems for NY – Long Lines and High Costs,” Report for New Yorkers for Verified Voting, 14 November 2006, pp. 1-10.