Using Quarterly Seasonal Adjustments to Improve Monthly Adjustments
David F. Findley, Brian C. Monsell
RR 93/11,10/18/93
ABSTRACT
Occasionally, time series which show weak or inconsistent seasonality when examined on a
monthly basis can be strongly seasonal when examined on a quarterly basis. An agency which
publishes such data on both a monthly and quarterly basis faces a dilemma: how should the
monthly series be seasonally adjusted to reflect the seasonality found in the quarterly series? In
this paper, we will examine what happens if the monthly seasonally adjusted series is
benchmarked to its quarterly counterparts. Since the benchmarking function to ensure that
months with erratic month-to-month changes are downweighted. This procedure is evaluated on
a number of Census Bureau economic time series, and the effect of the weighting scheme is
examined. Special attention is paid to the stability of the month-to-month change of the
seasonally adjusted series (as measured by the sliding spans diagnostics of Findley et.a. 1990)
and the amount of revision to the published monthly seasonally adjusted series caused by the
quarterly benchmarking.