The number of telephones in service in the U.S. at the end of each month from 1916 to 1942 show a seasonal slowdown in the telephone business in summer months. At the end of August, the number of telephones was typically 0.6% below trend.[1] As the telephone business picked up in the autumn, deviations from trend decreased until swinging positive in March. This seasonality is the net result of persons adding new telephone service and persons dropping telephone service. Given adds dominated drops to create a strong secular growth trend, the summer slowdown is plausibly interpreted as fewer businesses and residences installing new telephone services during summer months.[2]
Monthly toll service revenue from 1933 to 1982 doesn’t have a simple, easily interpretable seasonal pattern. The median deviation is largest in January: 3.2% below trend when monthly traffic is normalized to an equal number of days per month.[3] August does not show a summer-travel toll revenue spike, there’s no Mother’s Day May spike, nor a December holiday spike. Perhaps variation in residence and business toll revenue is negatively correlated across months and thus smooths aggregate toll revenue. Perhaps revenue reporting is not perfectly correlated in time with actual traffic. AT&T’ explicitly mentioned accounting for seasonality for its charts of the deviation of telephone traffic from normal from 1902 to 1929. January seasonality is not an effect that readily comes to mind. What exactly is going on with these data and toll-traffic seasonality remains unclear.
* * * * *
Graph: the graph shows box plots for the monthly sets of monthly deviations
Data: monthly U.S. data for telephones in service, 1915 to 1943, and for toll-service revenue, 1933 to 1982 (Excel version)
Notes:
[1] Specifically, -0.6% is the median percent deviation of August monthly data from the August-centered, 12-month moving average of the monthly number of telephones.
[2] Monthly data on the number of telephones in service is publicly available only from Oct. 1915 to June, 1943. Telephone companies know well the current pattern of telephone business seasonality. It would be interesting to know whether seasonality patterns have changed, and if so, why.
[3] Given that the number of days per month varies from 28 to 31 (11%), normalization is important. Normalizing with workdays rather than calendar days makes little difference to the pattern of median monthly deviations.