Generating monthly data from annual/quarterly data

[JanVinter]

I have a revenue dataset for various businesses. For about half of those businesses, monthly data is available. For the other half, only annual or quarterly revenue data is present. I know the seasonality of the revenues fairly well from the half of the businesses who have reported their monthly revenue. In addition, the data goes back at least five years, so I know that seasonal effects are real. What is the best way to go about interpolating monthly revenue for businesses with missing data given the information I have on hand? Even though it is not strictly true, I am assuming that the businesses with missing data and businesses with complete data are sufficiently similar that we need not worry about any underlying issues. I know that I can undertake frequency conversion of low frequency to high and select an option such as cubic spline interpolation, but is that going to identify and utilize the seasonal effects that are present in the data?

[CharlieEVIEWS]

As a suggestion, claiming only the vaguest, most minimal expertise in this field, you could estimate a SARIMA model from one of your monthly series and use the coefficients to interpolate your quarterly/annual.

[trubador]

No, such interpolation methods do not account for seasonality or any other type of low frequency fluctuations for that matter. However, new frequency conversion methods (Denton, Chow, Litterman) introduced in version 9 might be of help. These methods make use of an indicator variable when interpolating the missing values. Since you have monthly data for half of the variables, you can estimate the seasonal factors with regular adjustment methods (Census X12, TRAMO-SEATS). If these factors appear to be similar, you can average them to generate an indicator variable for seasonality. You can then use it during the frequency conversion...

[CharlieEVIEWS]

Or do what Trubador suggests, and avarage over all of them

[JanVinter]

Interesting. I will definitely give those new frequency conversion methods a try once I get access to EViews 9 on my work computer next week. They definitely appear to be a much better choice than cubic spline interpolation for my purposes.