DIALING BEHA VIOR OF SUBSCRIBERS 435 



tions that equalled or exceeded particular time intervals between the starts 

 of the first and second attempts. Figure 1(b) shows similar graphical data 

 for 211 Bronx-Westchester observations and Fig. 1(c) shows similar graphi- 

 cal data for 445 Brooklyn-Queens observations. Each of these three graphs 

 is compared with a composite curve for 1107 observations. This composite 

 curve is developed from the data on Fig. 2(a). 



Figure 2(a) shows, by dots, the cumulative percentage for 1107 observa- 

 tions, which are comprised of the 451 Manhattan, 211 Bronx- Westchester 

 and 445 Brooklyn-Queens observations, that equalled or exceeded particular 

 time intervals between the starts of the first and second attempts. A smooth 

 curve was drawn through these plotted data. This curv^e is also shown on 

 other figures, for the purpose of visual comparison of the various plots of 

 data with the overall results. 



Figure 2(b) shows a graph concerning 465 observations of the total 1107 

 observations. These are the cases where a busy was observed on a second 

 attempt. (Of the 1107 total observations, 817 resulted in a second attempt 

 within ten minutes and 290 were classified as abandoned. Of the 817 

 second attempts, 327 cases were able to complete their calls, 16 resulted in a 

 don't answer, 9 were referred to an operator and 465 encountered a busy.) 

 Figure 2(b) shows, by dots, the cumulative percentage of the 465 second 

 attempts that equalled or exceeded particular time intervals between the 

 starts of the second and third attempts. The graph of Fig. 2(b) does not 

 differ significantly from the composite curve for 1107 observations. This 

 feature indicates that, when observations concerning subscriber busies are 

 made, it is not necessary to have the first observed attempts coincide with 

 the first actual attempts. The observations can begin with any attempt. 



Figures 3 and 4 are graphs similar to that shown on Fig. 2(a), the dif- 

 ference being in the graphical ordinates used in order to present additional 

 pictorial representations of the data and to project the curve beyond the 

 observed limits. 



The percentage of subscribers who dial their calls again after encountering 

 busies is estimated from Figs. 3 and 4 to be 90%. The data on Fig. 3 are 

 projected to a time interval of 1,500 seconds (25 minutes). Judging by eye, 

 beyond this point, it appears that the curve is asymptotic to the 10% 

 horizontal line. This means that 10% of the subscribers abandon their 

 calls and 90% try again. The part of the curve on Fig. 4 that projects 

 beyond the limit of the observed data crosses the 10% line at 6,400 seconds, 

 an interval of Ij hours. This seems to be a very long time for a subscriber 

 to wait before redialing his call. It is unlikely that many attempts are made 

 beyond this period. 



Table III was prepared to determine the disposition of the calls on second 

 attempts and to see if a correlation exists between certain time intervals, 



