A Time Series Analysis of U.S. Inland Waterways Trade
Excel | CSV
"Owners, agents, masters, and clerks of vessels and other craft plying upon the
navigable waters of the United States, and all individuals and corporations engaged
in transporting their own goods upon the navigable waters of the United
States, shall furnish such statements relative to vessels, passengers, freight, and
tonnage as may be required by the Secretary of the Army" [from Section 11 of the
Rivers and Harbors Appropriations Act of 1922 (42 Stat. 1043), as amended, and
codified in 33 U.S.C. 555]. The monthly data reviewed here represent the tonnage
(in millions of short tons) of all commodities transported on the U.S. inland waterways
from January 1994 through March 2002.
Figure 1 - Transported Tonnage on U.S. Inland Waterways: All Commodities
Figure 1 provides the time series plot of the monthly tonnage. As is true of other
transportation data, the plot indicates strong seasonality in the data, with tonnage
dropping sharply in the winter months. The issue to be studied is whether there is
some significant direction to the trend underlying the seasonal variation. A cursory
glance gives the impression of some downward trend-does it really exist?
The next step of analysis is the measurement of the seasonality prior to the extraction
of the trend.
Figure 2 provides a plot of the monthly seasonal variation the transported tonnage
for inland waterway trade. The months of and February the lowest
months for freight movement; these two months are seen as the troughs the
actual data series (as see Figure 1). The series seem to around
October, which has the highest seasonal variation. The summer months are positive,
expected. December shows a negative impact on the overall monthly average,
but negative impact is not as strong that exhibited by January and February.
Figure 2 provides a plot of the monthly seasonal variation,for calculated over the period
of January 1994 through April 2002, for theJanuary transported tonnage for areinland
waterway trade. The months of January and February are the lowest months infor
freight movement; these two months seen arepeaksas inthe thetroughs in the actual occurdata
series (as seen in Figure 1). The peaks in the series seem to occur around October,
which ashas the highest positive seasonal variation. The summer months of July
and August thisare strongly positive, as expected. asDecember shows a negative impact
on the overall monthly average, but this negative impact is not as strong as that
exhibited by January and February. The monthly factors for March, April, June
and September are not significantly different from zero.
Figure 2 - Monthly Seasonal Variation of Transported Tonnage
To see how the seasonality compares to the seasonality in other modes of transportation,
we have provided a plot of the seasonal variation of highway vehicle miles
traveled in Figure 3 (the VMT analysis was originally provided in the September
2001 issue of Transportation Indicators).
Figure 3 - Monthly Seasonal Variation for VMT
The highway VMT seasonality shows a similar pattern to the water trade data.
Both experience large drops in the January and February months. For highway
travel, however, the highest months are July and August, with low autumn values.
Water tonnage movement, on the other hand, exhibits higher seasonality for October
and November; the summer months are closer to the overall average.
The long-term pattern of the tonnage data tends to be masked by this strong seasonal
pattern. The trend can be extracted as a separate component, so that long
term behavior can be studied. Figure 4 provides a plot of the underlying trend
compared to the actual monthly data. The long-term slope to the trend proved to
be insignificant, which is confirmed by the manner in which the underlying trend
tends to drift up and down over the period of seven years.
In the earlier data in this time series plot, a strong intervention was found around
May 1995; this particular data point is highlighted in Figure 4. This intervention
was removed prior to fitting the final trend line.
Statistical process control (SPC) procedures were applied to the model to search
for additional shifts in the data. The SPC also highlighted a significant decrease in
the data that could not be explained by seasonal variation alone. As the seasonal
variation showed in Figure 2, December 2000 seasonality tends to be lower than
the average monthly value. In fact, the trend shows a rise early in the year, followed
by a significant decline in the last few months of 2000; this temporary decline
may have been due to the adverse impact of the extreme heat, drought, and
other disasters on agricultural products that year.
Figure 4 - Transported Tonnage on U.S. Inland Waterways and Underlying Trend
|
