Time series analyses rely on consistent temporal and spatial reporting units, but these units have not been consistent over the entire period. In the beginning of 1988 the reporting frequency changed from monthly to quarterly, a change which can be readily addressed through a variety of established statistical techniques. A more significant statistical challenge is Timber Mart-South's change in 1992 from (typically) three reporting regions per state to two. We developed conversion tables to address this change in reporting areas, permitting longitudinal analyses of the current two regions per state but extending back to the beginning of Timber Mart-South's reports in 1976.
The conversion figures provided assign product prices from a discontinued area to each of the two remaining areas in the state based on county-level timber removals. Timber removals were obtained from the Forest Inventory and Analysis Eastwide database of surveys ending between 1984 and 1991 which report average annual removals over the preceeding 6 to 10 years. The conversion weights represent the ratio between removals from counties in the new area's portion of an old area to removals in the entire new area to which they were reassigned.
SOURCE and SCOPE: The data span from Arkansas, Oklahoma and East Texas through Kentucky, Virginia and Florida (inclusive) and may be applied to TMS' prices reported from December 1976 through 1991. The conversion weights are applicable to southern yellow pine and mixed hardwood prices for sawtimber and pulpwood stumpage, and pulpwood and sawlogs delivered to the mill.
As with any aggregation change, weighting schemes introduce errors. We have used rigorous statistical tests to evaluate the seamlessness of our weighting method, tests which are described in an online presentation and a paper in press. The results show that the conversion weights produce seamless time series for the vast majority of the products and regions examined. We therefore conclude these weights can be applied with some confidence in statistical analyses.
RECORDS and VARIABLES:
state ="State name" period ="years over which removals were calculated, used to calculate weights" new area ="TMS reporting area beginning in 1992" old area ="TMS reporting area prior to 1992" weight ="weight applied to pre-1992 prices for aggregating to new"
EXAMPLE 1: For the period December 1976-Quarter 4, 1991, Alabama had three (old) Timber Mart-South geographical Areas. Price reporting regions were redrawn in 1992, dividing the state into two new Areas. Hence, parts of each old Area now fall into each new Area. Suppose a statistician wanted a consistent series of southern yellow pine stumpage prices for both new Alabama reporting regions from 1988, Quarter 1, to 1998, Quarter 3. For the period 1992-1998, two prices have been reported per state but for 1988-1991, three prices were reported. Suppose that the prices for 1988 Quarter 1 were as follows:
Old Alabama Area 1: $110/mbf
Old Alabama Area 2: $140/mbf
Old Alabama Area 3: $190/mbf
The new Area 1 and Area 2 prices can be computed by using the southern pine sawtimber conversion weights. The equivalent new Area 1 price corresponding to that quarter is therefore:
(0.4905 * $110/mbf) + (0.5095 * $140/mbf) + (0.0000 * $190/mbf) = $125.3/mbf.
The equivalent new Area 2 price corresponding to that quarter is then:
(0.01088 * $110/mbf) + (0.4427 * $140/mbf) + (0.5385 * $190/mbf) = $166.4/mbf.
EXAMPLE 2: Mississippi prices were reported for three Timber Mart-South Areas from 1976-1991 and for two from 1992-present. Suppose that mixed hardwood pulpwood delivered log prices in May 1980 were as follows:
Old Mississippi Area 1: $27/cord
Old Mississippi Area 2: $25/cord
Old Mississippi Area 3: $33/cord
The new Area 1 and Area 2 prices can be computed by using the Mississippi hardwood pulpwood conversion weights. The equivalent new Area 1 price is:
(0.1373 * $27/cord) + (0.5335 * $25/cord) + (0.3292 * $33/cord) = $27.9/cord.
The equivalent new Area 2 price is:
(0.0000 * $27/cord) + (0.0000 * $25/cord) + (1 * $33/cord) = $33/cord.
This last example highlights an important issue: some new Areas are wholly contained within one old Area.
created by: Jeff Prestemon