This paper proposes a new method for estimating a monthly regional pro
duction model. The technique involves treating the region's monthly in
dustrial output as a latent variable, which is in turn a function of c
apital (proxied by energy usage) and labor inputs. Annual observations
on regional value added correspond to the summation of the unobservab
le monthly series over the 12 months, while changes in the national In
dustrial Production index help infer the series' month-to-month fluctu
ations. The model is estimated using the Kalman filter and the method
of maximum likelihood. The estimates are used to compute monthly indic
es of regional value added for 15 individual 2-digit industries, and f
or the aggregate manufacturing sector in the Seventh Federal Reserve D
istrict. In a comparison of out-of-sample forecasting accuracy, the mi
xed-frequency model outperforms both the traditional parametric Cobb-D
ouglas and nonparametric Atlanta methods over the 1988-89 forecasting
horizon.