Seasonal Adjustment for Texas Manufacturing Outlook Survey

The Dallas Fed began reporting its Texas Manufacturing Outlook Survey (TMOS) on a seasonally adjusted basis with the August 2009 release. The change enhances the survey’s value by making it easier to discern underlying economic trends.

Time series can exhibit seasonal patterns that, if not accounted for, can lead to misinterpretations about real economic activity. However, seasonality cannot be statistically identified without at least three years of data. The Dallas Fed began conducting the manufacturing survey in June 2004. By August 2009, it had accumulated 62 months of data, enough to make precise seasonal adjustments.

The Dallas Fed uses the X12 seasonal-adjustment procedure, developed by the U.S. Census Bureau, to statistically remove seasonal effects.  X12 requires a minimum of 36 months to seasonally adjust a time series, although 60 months is preferred for a high degree of confidence in the statistical tests.

TMOS respondents are explicitly asked to take seasonal variations into account in assessing their firm’s performance each month. However, the initial X12 results indicated that 17 of the 34 indexes contained statistically significant seasonality.

As a result, the following indexes have been seasonally adjusted:

• current production
• future production
• current capacity utilization
• future capacity utilization
• current volume of new orders
• current growth rate of new orders
• future growth rate of new orders
• current unfilled orders
• current volume of shipments
• future volume of shipments
• current prices paid for raw materials
• current prices received for finished goods
• future wages and benefits
• current number of employees
• current company outlook
• future company outlook
• current business activity

Each index is calculated by subtracting the percentage of respondents reporting decreases in activity from the percentage reporting increases.

Indexes were tested using the X12 procedure’s F-test for seasonality assuming stable seasonal factors.  When no seasonality was detected, the indexes and their components were not adjusted. For those series identified as seasonal, the increase, decrease and no change components were each adjusted.  The index was then re-computed using the adjusted components. If the three adjusted component series didn’t sum to 100 percent, they were normalized to add up to 100. 

In January each year, the Dallas Fed will revise historical data for the Texas Manufacturing Outlook Survey by recalculating the seasonal adjustments.

Notes:

1. For a more detailed description of seasonal adjustment, see DataBasics "Seasonally Adjusting Data."
2. For reference purposes, the nonseasonally adjusted data will continue to be available.