10-7-25 It sure doesn’t help forecasters when we can’t get fresh data. So how can we forecast when there is no current data forthcoming?

One of the data sets that my forecasting models are set to calculate is annual forecast on smaller sets of data within the year. For example: The Total Spending SAAR average for the months of AMJJAS (Apr thru Sept) when extended for a full 12 months, predicts the annual spending within +/- 1%. This calculation has been within this 1% limit 22 times in 24 years (think Drake Maye accuracy here). The other 2 years were off by 1.1% and 2.2% (in 2020). And unfortunately, we don’t have August or September data.

The last actual data we have is July. For Nonresidential Bldgs, the average predicted from AMJJ (Apr thru July) predicted the annual total Nonres Bldgs spending within 1.8% 13 out of the last 14 years (2011 to 2024), with one year being 2.6%. The average of those 14 years is less than 1% off from the actual.

This is not a simple average. Each month produces a different share of annual spending. For ex: Based on 20-year averages, January and February each produce only about 7% of annual Nonres Bldgs spending while June and July produce each about 9% of annual spending. When a small set of data is used, the spending in each of those months is compared to the 20-yr average for those months and projected out to get 100% or a full 12 months of spending.

The small data set of AMJJ, when used to predict 2025 Nonres Bldgs spending, forecasts +/- 0.9% or $730bil to $756bil. The average predicts spending on Nonres Bldgs will hit $743bil. My current forecast, which includes actual data thru July and forecast to year end, predicts spending will hit $749bil.

Residential spending, based on small data set is between $882bil and $920bil., with the average predicting $901bil. My current forecast with all actual data from Jan thru July and forecast thru year end is $892bil.

The best small data set predictor is having data for AMJJAS. This predicts total spending will come in between $2109bil and $2149bil, with the average for the year at $2133bil. My current forecast predicts spending for 2025 will total $2140bil.

Generally, I look at these small data set calculations to insure I haven’t made some kind of blunder in my forecast. In this case, the small data set gives some comfort level that the annual forecast, although based on only 7 months of actual data, may not be very far off from what to expect at year end.