College of Business Administration
University of Rhode Island
William A. Orme
WORKING PAPER SERIES
encouraging creative research
Is Freight Rate Relevant to Port Volume Forecasting?
Evidence from the Guangzhou, China Port
Huanxin Zhang, Xia Pan, and Jeffrey E. Jarrett
2013/2014
This working paper series is intended to
facilitate discussion and encourage the
exchange of ideas. Inclusion here does not
preclude publication elsewhere.
It is the original work of the author(s) and
subject to copyright regulations.
No. 7
Office of the Dean
College of Business Administration
Ballentine Hall
7 Lippitt Road
Kingston, RI 02881
401-874-2337
www.cba.uri.edu
Is Freight Rate Relevant to Port Volume Forecasting?
Evidence from the Guangzhou, China Port
Huanxin Zhang
Lingnan College，Sun Yat-sen University, China
Xia Pan
Lingnan College，Sun Yat-sen University, China
Jeffrey E. Jarrett
Management Science
University of Rhode Island
Is Freight Rate Relevant to Port Volume Forecasting? Evidence
from the Guangzhou, China Port
【Abstract】
We study historical data of the cargo going through the Guangzhou (GZ) port and
related research the relationships between cargo shipments through the GZ port and its
relation to domestic and international shipping prices (rates). In turn, we develop a
regression based forecasting model based on the data of the GZ cargo port. The second
task is to introduce the GZ port, the international dry bulk shipping market; the Chinese
Coast Bulk freight index (CCBFI); and the Baltic Dry Index (BDI) which reflect
domestic and international freight rates respectively. The third task is to make use of the
data of the GZ cargo port, CCBFI and BDI from January 2004 to February 2010. The
developed model establishes a multi-linear regression to relate the impact of the
previous month BDI and CCBFI on the current GZ port cargo and determine the
magnitude of the effect. Second, we establish a time series-regression forecasting
model. This requires us to observe and consider including historical data of BDI,
CCBFI and GZ cargo and come to a conclusion that relates the impact of BDI and
CCBFI on the GZ cargo port. Finally, by developing a two parameter exponentially
weighted moving average (EWMA), we obtain forecast with high predictive accuracy
【Keywords】Port Handling Capacity; Dry Bulk Freight Rates; EWMA; CCBFI;
BDI;
Introduction
In this paper, we analyze the data of the GZ cargo port cargo, CCBFI and BDI
from January 2004 to February 2010 by using a variety of models. There are two main
research purposes. First, through data analysis, we study the relationships between
Guangzhou cargo port and domestic and international shipping prices. Second, by
establishing a Guangzhou cargo port cargo forecasting model, we obtain an accurate
prediction of the magnitude of the activity of the Guangzhou cargo port cargo
The GZ Port is a window of foreign trade for this region of the People’s Republic
of China (PRC). The cargo port is a weathervane of the international trade activity level
of the port and its regional environment. Through the forecast of the cargo port activity,
we, in addition, also determine the trends of international trade in the Pearl River Delta.
For the long term, predicting the cargo port activity with great accuracy is of great
importance in designing a reasonable and scientific port layout; the basic scale of
investment, business strategy; and developing strategies for integrated transport
planning.
1.1 Previous research
The cargo port activity is statistically a time series with random fluctuations which
is affected by political, economic and natural and other factors. Cargo port activity
forecasting systems are complex. There is not a universal technique applied for this
system up to now. More commonly, we use linear regression, nonlinear regression, time
series prediction method, logistic function fitting method, exponential curve fitting
method, gray system prediction method and elastic coefficient method and many other
methods to solve this system.
Currently, the researches on the port throughput forecast focus on three areas (Sun,
2010; and Bakshi, Panayotov and Skoulakis, 2011).
First, making local improvements on the existing prediction methods, so we can
obtain the prediction result which is closer to the true value. By adjusting the inflection
point of the Logistic curve, researching the methods of blur compensation and
estimation of optimism and pessimism on production trends including Logistic curve
prediction methods.
Second, some experts predict the cargo port handling activity by referring to
methods from other applications including the use of econometrics, control theory and
related methods Others have produced dynamic PHQDF models to predict the cargo
port activity, which systematically combines the historical trend of cargo activity, the
socio-economic development of the hinterland, politics, policy, and psychological and
technical factors. The example tests show that the model has higher prediction
precision and better prediction results than the models suggested above.
Third, have the raw data processed. And through selecting appropriate explanatory
variables from a plurality of explanatory variables, we may obtain better forecast
results. Regression models to predict the cargo port activity are not uncommon. As
results often indicate, when one utilizes mathematical models to predict, the most
critical aspet is that the accuracy of the model, relationship of the response variable and
predictors variables and the quality of data is most important. The complexity of the
model, especially in the case of only having small amounts of historical data is of lesser
importance. Some examples of forecasting port activity include Wang, 2006, Xu, Yan
and Zganh (2006), Hui , Seabrooke and Woong (2004), Yuan and Xie (2004), Chen,
Zhu, and Yu (2005), Cheng and Wong (1997), Lam, Pan, et al. (2004), Sun and Wang
(2007), Sun and Zheng (2007), Yang, and Guan (2005, and Sun (2010. In each
application, they use very large amounts of data to predict port capacity, tonnage and
other measures of activity.
In this paper, by using large amounts of historical data and establishing
multi-linear regression model and time series regression model, we study the impact of
the domestic and international shipping freight on the current GZ cargo port activity the
In addition. We construct a two parameter exponentially weighted moving average
model (EWMA), we develop a better model for forecasting activity for the a major port
in the PRC.
GZ Port and dry bulk shipping market; Overview
According to the Harbor Law of the PRC, a port refers to a region consisting of
landside area and waters and with the functions as ships entering and leaving, mooring,
berthing, passengers leaving and entering the ship, cargo handling, storage and other
functions containing the corresponding terminal facilities. Ports are the assembly point
and hub of the water and land transportation, and also they are collecting and
distributing centers of the industrial and agricultural products and the import and export
goods. The port is a natural interface linking the hinterland with maritime transport, and
as a result, ports are a special node of international logistics.
According to the locations, ports can be divided into the estuary ports, harbors and
river ports . Estuary ports are located in estuaries or the segments of tidal estuaries and
can also serve for both seagoing vessels and riverboats. Large cities are the basis for
convenient transportation. The inland waterways are often extend to the vast economic
hinterland, undertaking a lot of goods traffic. As a result, many of the ports around the
world are built near the estuary, such as the Ports of Rotterdam, London, New York,
Leningrad, and Shanghai. Estuary ports are characterized by terminal facilities
arranged along the river bank, the distance between the port and the sea is short and do
not need to build breakwaters. Besides, if the length of shoreline is not enough, you can
add the excavated basin. GZ Port belongs to the estuary port grouping.
Harbors are located on the coast or in the bay or lagoon, but also some are built on
the deep sea which is far from the coast. The ports located in the open sea shore or in the
bay lacking of natural reserve ports are usually required to build the considerable size
breakwater, such as the Port of Dalian, Qingdao Port, the Port of Lianyungang, the Port
of Keelung, the Port of Genoa and others. The single point or multi-point mooring
docks served berthing for supertankers and ore ships belong to the harbors off the coast
having no screenings, such as Marsa el Brega (Libya), the Port of Sidon (Lebanon) and
others. In addition, there are also large seaports rely entirely on natural cover, such as
the ports of Tokyo, Hong Kong, and Sydney.
River ports are located on the natural rivers or canals, including lake harbors and
reservoir ports. Lake harbors and reservoir ports have broad surface and sometimes
large waves. Therefore, there are many similarities with the harbors, for example, they
often need to build breakwaters. The ports built on the Kuibyshev, Tsimlyansk and
other large reservoirs in Russia, St. Louis, Memphis and Philadelphia in the United
States and the small ports built on the Hongze Lake in the PRC belong to the river port
group.
Ports, especially the coastal ports, make a good connection between hinterland and
overseas transportation. With economic development as a goal and international
division of labor, these ports play an important role in the development of the
hinterlands in many aspects. A Port is a junction point of sea and land transportation
where a combination of the waterway transport, road transport and rail transport merge.
These ports achieve the ability of multimodal transport. Goods coming from other
countries or regions will be transported by road transport to inland areas. Also, a Port is
an industrial base designed to reduce transportation costs of raw materials or finished
products, many manufacturing companies including steel industries, petrochemical
industries, shipbuilding industries and other heavy industries have chosen to set up
factories in the surrounding area of the port to carry out industrial activities. In addition,
the enterprises processing imported materials then exporting also tend to set the plant at
the bonded area around the port. The Port is an integral part of the “Supply Chain,”
from raw materials to end user products.
The Port is an integrated logistics center with the purpose of building international
logistics nodes. The ports carry out activities including not only the traditional logistics
warehousing and transport operations but also customs clearance, freight forwarding,
shipping agency, buyer groupage, distribution processing, container leasing and other
activities.
Excellent ports provide cities with a window for external development, in which a
city makes full use of its own superior resources in the international division of labor to
gain profit from the international trade and promote urban development. Developing
cities with their harbor and port development is not a new idea, but is a mature method
for creating wealth and the movement of goods and services.
Port handling capacity refers to the total amounts of goods imported and exported
to the port by water transport and after handling operation in a certain time. The unit of
the dry bulk cargo throughput is ton. The unit of container port is TEU (Twenty-foot
Equivalent Units). Port handling capacity is the important quantity index of the results
of the operating activities reflects of the port and the size of the port. The flow structure,
quantitative structure and physical classification structure of the port handling capacity
are the most direct expression of status, role and impact at the international and regional
water transport chain, but also a statistical measure of national, regional and urban
construction and development.
Factors that affect the port handling capacity are complex and can roughly divided
into two parts; one is objective regional factors, such as the size of the hinterland, the
level of production development, export-oriented economic development and the
quantity of import and export commodities. The other portion refers to its own
conditions for constructing harbors which include natural conditions and
socio-economic factors. When we obtain the above conditions the, the level of labor
organization and management, the quantities of handling machinery, technical level,
ship type, vehicle type, hydrological and meteorological conditions, seasonal
characteristics of the industrial and agricultural production, the balance of ships and
vehicles travel to the port, as well as the varieties and quantities of goods through the
port handling are all likely to become important factors affecting the port handling
capacity. The most direct and critical factor is the size of the ability of berth, that is,
the maximum tonnage limit of the port for berthing ships.
GZ Port Development
GZ Port is located in the Pearl River estuary and the center of the Pearl River Delta
where the export-oriented economy is most active for PRC China. It is near the South
China Sea, adjacent to Hong Kong and Macao. The East River, West River and North
River flow into the sea at this point. Through the water system of the Pearl River, the
GZ Port communicates with major cities in the Pearl River Delta, i.e., Hong Kong and
Macau, and meeting the southwest of PRC by the West River. Through the seaward
channel of the Ling ding Sea, the GZ Port can communicate with coastal harbors in the
PRC and the harbors all over the world. The geographical location of GZ is shown in
Figure 1.
_________________________________________________________________
Figure 1 Geographical location map of GZ Port
________________________________________________________________
GZ Port has a long history and is an important port for ancient Chinese foreign
trade, one of starting points of "Marine Silk Road". In Tang-Song period, "Guangzhou
Marine Silk Road" is the world's longest ocean routes. In Qing Dynasty, GZ Port
became the only external port and the largest foreign trade port. Since China carried out
reform and open policy, the rapid socio-economic development make GZ Port an
important hub of integrated transport system and an important foreign trade port in
southern China. It is one of China's container shipping transport arteries and the largest
bulk cargo (energy, raw materials, etc.) transshipment port and container transportation
trunk port in southern China.
GZ Port is divided into the Inner Harbor, Huangpu Port, Xinsha Port, Nansha Port
and the Pearl River estuary. The inner harbors based on the original eight inland
waterways of GZ mainly offer energy, raw materials, food, bulk cargo and container
handling and passenger transport services for the region of GZ and the Pearl River
Delta. The Huangpu Port mainly undertake the coastal and offshore, container
transportation and the bulk cargo transport including food, coal, chemical fertilizer,
refined oil. The Xinsha Port focuses on container, coal, ore, petrochemical, food,
chemical fertilizer and other materials transport. Both the Nansha Port and the Pearl
River Estuary water area are the integrated ports, the work area of Shazai Island focuses
on the Ro-Ro and general cargo transportation, the main work of Xiaohu operation area
is the energy, liquid chemical transportation, Lu Wan operation area mainly works for
general cargo transportation, Nansha operation area dominated by foreign trade
container transportation develops the bonded function, logistics function, commerce
function and other functions, and develops and undertakes large bulk cargo
transportation combining harbor industry.
GZ Port has a number of advanced facilities such as the large containers and the
specialized deepwater docks for coal, grain, oil and chemical, as well as the largest
ro-ro ship pier in the south region of China. As of 2008, Guangzhou Port has 631 quay
berths of all kinds, 88 anchorages, 23 buoys, 59 big tonnage berths; and the maximum
tonnage is 300,000 tons.
GZ as a transportation hub in southern China has developed waterways, rail and
road transport. There are crisscrossing waterways in the Pearl River Delta. Shipping
resources are in excellent condition. The harbor district has five special railway lines
where
Beijing-Guangzhou,
Beijing-Kowloon,
Guangzhou-Shenzhen,
Guangzhou-Zhanjiang and Guangzhou-Meizhou-Shantou railway intersect. Highways
extend in all directions. Highways of each port connect with expressways, highways
and national highways. "Nansha Port Express" through to the Nansha harbor district.
On the basis of the sound infrastructure of land and water transportation, the hinterland
of GZ Port not only cover the Pearl River Delta which heart was Canton but also extend
to the Pan-Pearl River Delta region, which make the circulation of goods more
substantial.
Entering the 21st century, on the basis of sustained and rapid economic growth of
the hinterland, GZ Port developed rapidly. Beginning in 2004, for three consecutive
years, the growth of the cargo throughput of the GZ port is at the speed of 50 million
tons per year. In 2009, cargo throughput of GZ Port is 364 million tons, ranked fourth in
the country and the container throughput is 11.2 million TEUs, ranked third in the
country and sixth in the world. Two main indicators are among the world's top ten ports.
At present, GZ Port connects with more than 350 ports in more than 80 countries and
regions around the world and more than 100 domestic ports. GZ port cargo in the last 10
years is shown in Figure 2.
_____________________________________________________________
Figure 2 Histogram of the Guangzhou port cargo from 1999 to 2009
______________________________________________________________
Dry bulk shipping market and major freight index
Narrow shipping markets refer to all kinds of shipping exchanges such as the
London Shipping Exchange, Baltic Exchange, Shanghai Shipping Exchange and others.
Generalized shipping market refers to the trading relationships of the shipping industry.
Transaction object is shipping products, namely all the services that is offered by
shipping enterprises for the shipping market to meet the needs of the displacement of
people and goods to and connect with shipping tools and equipment. According to the
shape and packaging of goods, the transported goods across oceans are classified as
three categories including liquid cargo, dry bulk and general cargo by maritime sector.
Liquid cargoes include oil, refined oil, liquefied gas, liquid chemicals and other liquid
cargo. Dry bulk includes a variety of primary commodities, raw materials. On the basis
of the size of bulk transport, bulk cargo is divided into two types as large bulk cargo and
minor bulk cargo. The large bulk cargoes mainly include coal, metal ores and grain.
Minor bulk cargoes include steel, wood, chemical fertilizer, cement, etc. General
cargoes
mainly
include
mechanical
and
electrical
equipment,
chemicals,
pharmaceuticals and other light industrial products, farm products and pasture products,
fishery products and so on. Dry bulk shipping market refers to the trading relationships
of all the services that is offered by shipping enterprises for the shipping market to meet
the needs of the displacement of dry bulk cargoes among different countries and
regions and connect with shipping tools and equipment.
The relation between supply and demand in dry bulk shipping
markets
Dry bulk shipping markets include two main elements, the demand for dry bulk
shipping and dry bulk shipping capacity. Demand for dry bulk shipping refers to the
quantity of dry bulk cargoes transported by sea among countries and regions at a certain
period of time, which is a demand factor of the shipping market. Dry bulk shipping
capacity refers to the total transport capacity of dry bulk carriers including Handymax
bulk carriers, Panamax, Capsize bulk carriers and the Lake bulk carriers in the current
dry bulk shipping market, which is a supply factor of the shipping market.
In the dry bulk shipping market, if dry bulk shipping capacity cannot meet the
demand for dry bulk shipping, dry bulk shipping freight rates will rise. Meanwhile, if
dry bulk shipping capacity significantly exceeds demand for dry bulk shipping, dry
bulk shipping freight rates will decline. In addition, commodity price changes result in
changes in supply and demand. Holding other factors constant, price rise leads to
reduced demand and prices fall leads to increased demand. Similarly, price rise causes
increased supply and price fall leads to reduced supply. Specific performance in the dry
bulk shipping market, the rise in dry bulk freight rates will lead to a decline in demand
for dry bulk shipping, while the fall in freight rates will promote the rise in demand for
dry bulk shipping. In contrast, the rise in dry bulk freight rates will lead to the rise in dry
bulk shipping capacity, while the decline in freight rates will promote a decline in dry
bulk shipping capacity.
Price fluctuations by adjusting supply and demand, so that tends to balance, but in
the shipping market, because demand for shipping derived resistance, potential, and
long life cycle marine, construction costs are high, long construction period and other
reasons, often entice shipping bullwhip effect occurred in the market, resulting in the
shipping market supply and demand imbalance.
For Chinese ports, being able to reflect the dry bulk shipping market there are two
tariff indexes, respectively, to reflect the international shipping market, the Baltic Dry
Freight Index, as well as dry bulk shipping market reflecting domestic freight China
coastal (bulk) freight index.
Baltic Dry Index (Baltic Dry Index, abbreviations BDI, [see Bakshi, Panyatov and
Skoulakis, 2010], is an index of economic activity in the shipping industry, including
dry bulk shipping industry, trading volume changes, known as the international trend of
the overall dry bulk market barometer. BDI is composite index being composed of
Capsize, Panamax and Handysize vessels, each occupying third of the weight of the
index. It was weighted according to the spot freight of twelve major routes to reflect the
current market price. Therefore, the change of shipping price will result in a change in
the index. For the index to be fair its manager, the Baltic Exchange, has very strict rules
on calculating this freight index. These rules include:
(1) The selection of routes requires geographical balance, the route reflects the
trade of both the Atlantic and Pacific, as well as the inter-oceanic trade (Maintaining the
balance of round-trip routes). The weight of each route does not exceed 20%
(2) The routes including in the calculation of the index have a certain turnover, or
important relevant routes. Seasonal routes are not considered.
(3) There is a reasonable amount of accurate transaction reports. The routes
potential or actual rented or controlled by one or a few charterers are not considered.
(4) 20 brokerage firms which are internationally renowned, reputable,
representative, consisting of three groups, each responsible for calculating the day of
the ship freight index.
China Coastal Bulk Freight Index (abbr. CCBFI) as a "barometer" of the coastal
shipping market, reflects the trends of price movements of coastal shipping market
without delay and help shipping companies, shippers, trade enterprises, ports, agents
and other related businesses gain market information and grasp the market dynamics.
(1) The base period. China coastal bulk freight index takes January 2000 as the
base period. The base index is 1000 points.
(2) Sample route selection. Based on the importance of the selection, it takes top
five kinds of goods in our coastal port cargo throughput as samples of goods of coastal
freight index, including coal, crude oil, refined oil, metal ores and grain. Based on the
volume size, taking regional coverage and future trends of routes into consideration, it
selects 21 samples of domestic coastal routes.
(3) Tariff information collection. Currently, there are 27 major incoming units to
provide tariff information.
(4)Release mode. Shanghai Shipping Exchange prepares and publishes China
Coastal Bulk Freight Index, sub-indices and the index of 21 routes every Wednesday.
GZ Cargo Port Forecasting Models
This section utilizes three models to predict GZ port cargo activity by making use
of the collected historical data of the GZ cargo port freightage and domestic and
overseas dry bulk freight. First, taking the last BDI and CCBFI as the explanatory
(predictor) variables and the current GZ port activity as response variable, we build a
multi-linear regression model to study the relationship of these two indices on the GZ
port activity. In turn, we introduce the time series of GZ cargo port to build a model to
predict the GZ port activity, test the validity of the model and estimate the influence of
dry bulk freight activity. Finally, using the historical data of GZ port activity, we aim at
building a forecasting model having predictive accuracy by making use of the
two-parameter exponential smoothing method. The multi-linear by OLS regression
results is the following;
Yt 1  0  1 X1t  2 X 2t
Where,
Yt 1  2626.134  0.0366 X1t  0.044 X 2t
Yt+1 = period ahead prediction of GZ cargo port activity,
X 1t = BDI, the mean monthly value
X 2t = CCFBI (mid-monthly value)
Table 1 contains the output of EVIEWS® (multi-linear regression software). [For
complete discussion of Regression modeling see Fahrmeir, Kneib, Leib and Marx,
2013.]Observe the coefficient of determination, R 2 = .0196 and the adjusted
2
coefficient, R  0.0088 .These coefficients indicate that the regression of port
activity is largely unexplained by the multi-linear regression on the variation in the
two freight indexes, BDI and CCBFI. In addition, significance tests would not reject
the hypothesis that there is no relation between the response variable and the two
predictors, BDI and CCBFI.
In addition, Table 1 indicates that tests for the significance of the individual
predictors result in not rejecting the hypotheses that each predictor taken individually
is related to the response variable. The p-values for both t-tests are large, that is,
p  value   .
The output is shown in Table 1 based on 73 sets of corresponding data.
_________________________________________________________________
Table 1 Output of the multiple regression analysis
Variable
Coefficient
Std. Error
t-Statistic
Prob.
0
2626.134
349.2942
7.518401
0.0000
X 1t
2626.134
0.047248
0.774825
0.4411
X 2t
-0.043952
0.300456
-0.146285
0.8841
R-squared
0.019585
Mean dependent var
2718.985
Adjusted R-squared
-0.008833
S.D. dependent var
547.8405
Durbin-Watson stat
0.349792
Prob(F-statistic)
0.505416
_________________________________________________________________
Figure 3 The residuals, actuals and fitted values are shown in one scatterplot.
_________________________________________________________________
The residuals indicate that there may be a trend in their values. This appears
corroborated by comparison of the actuals and fitted values. Hence, one may conclude
that improvements can and will be made to the forecasting model.
In summary, the previous monthly mean values of BDI and the mid-monthly
values of CCBFI may influence the current GZ port handling capacity. However, other
factors not included in the multi-linear regression probably influence the GZ port
handling capacity. These factors may have a time dimension.
In the linear regression model, we failed to take into account that international dry
bulk shipping prices and domestic coastal bulk shipping price as influencing factors.
Shipping rates for both international and domestic cargo due tend to increase over time
especially for the shipping industry because rates are reflected in long term contract
between the shippers and haulers. One method is to include a surrogate variable for
time. Hence, by including a time series component in the forecasting model perhaps
will improve our forecasting model .
In the next section, we consider the use of combinations of time series and
regression analysis model.
Autoregressive Forecasting Model
An objective of time series methods is to discover a pattern in the historical data,
develop a forecasting model and then extrapolate the pattern into the future. Time series
method is often used to provide predictions of future by taking advantage of the
chronological data. Autoregressive prediction model refers to the establishment of a
regression equation for prediction by making use of the existing association in the
individual observations of an historical time series. Regression and time series model
refers to the model combining of the time series analysis and regression analysis and
taking the influence of the time series and other factors on the time series into account.
In this application the modeler selects the predictor variables and may express the
model as follows:
Zt   X t  Yt  0  1Z( t 1)  2 Z( t 2)  3Z( t 3) 
 n Z( t n )
On the basis of the 74 sets of data in Appendix A, we make use of time series
analysis and eliminating several sets of data in January 2004, we analyze the residual
values including the monthly mean value of BDI, the mid-monthly value of CCBFI and
the monthly cargo handling capacity of the GZ port.
We use Z t to identify the response variable of the GZ port cargo handling
capacity, X t 1 ， X t 2 ， X t 3 … to identify the monthly mean value of BDI of each
period and Yt 1 ， Yt 2 ， Yt 3 … to identify the mid-monthly value of CCBFI of each
period. In turn, the multiplicative autoregressive model is expressed as follows:
Zt  1 X (t 1)   2 X (t 2)  3 X ( t 3) 
  2Y(t 2)  3Y(t 3) 
  j X (t  j )  1Y( t 1)
  mY(t m)  1Z(t 1)  2 Z(t 2)  3 Z(t 3)  n Z(t n )  
Where j, m, n  t j, m, n  Z .
For changing values of j, m, n , we present the outputs of R 2 in Appendix B.
R 2 is maximum value when j  1, m  2, n  6 .
When j  1, m  2, n  6 , the value of R 2 get to maximum, R 2  0.588 . The
outputs are shown in Table 2. The regression equation is
Zt  0.0204 X (t 1)  0.0423Y(t 1)  0.2535Y(t 2)  0.5618Z (t 1)  0.0201Z (t 2)
0.0414Z (t 3)  0.0747 Z (t 4)  0.0996Z(t 5)  0.1618Z(t 6)  757.015
[Fore greater information of multi-linear regression time series modeling see
Pindyck and Rubinfeld, 2010.]
________________________________________________________________
Table 2 Output of analysis on Regression - Time series forecasting model
Variable
Coefficient
Std. Error
t-Statistic
Prob.

757.0150
319.8731
2.366611
0.0214
X ( t 1)
0.020419
0.028933
0.705738
0.4833
Y( t 1)
0.042297
0.251350
0.168278
0.8670
Y( t 2)
-0.253484
0.220607
-1.149029
0.2554
Z ( t 1)
0.561794
0.139947
4.014328
0.0002
Z ( t 2)
-0.020117
0.166008
-0.121181
0.9040
Z ( t 3)
-0.041358
0.16967
-0.247700
0.8053
Z ( t 4)
0.074706
0.166727
0.448077
0.6558
Z ( t 5)
0.099610
0.169451
0.587841
0.5590
Z ( t 6)
0.161814
R-squared
0.644852
Mean dependent var
2813.638
Adjusted R-squared
0.587775
S.D. dependent var
479.5500
Durbin-Watson stat
1.884267
Prob(F-statistic)
0.000000
0.141661
1.142267
0.2582
The multiple coefficient of determination R2  0.645 . The adjusted multiple
coefficient of determination R 2  0.588 . The Durbin-Watson Statistic is 1.884
indicating that there is no evidence of autocorrelation in the residuals. The F-test for
overall regression indicates that for the regression, taken together the predictor
variables are related to the response variable. Further, the scatterplots of the residuals ,
actual and fitted values are all shown in Figure 4.
____________________________________________________________
In Figure 4 Scatterplots of residuals, fits and actual values
________________________________________________________________
We observe that there is no trend or pattern in the scatter of the residual
which corroborates the results of the Durbin-Watson statistic. Also, the
patterns in the fits and actual observations indicate that the forecast
contain a much smaller amount of error (magnitude of residuals) than in
the first multi-linear regression model.Application of the forecasting
model. Hence in Table 3, we observe the implementation of the model in
predicting cargo handling capacity.
Table 3 Partial Data Table of BDI, CCBFI and cargo handling capability in 2009
月份
BDI
CCBFI
Volume（in 10k tons）
2009年9月
——
——
3616.29
2009年10月
——
——
3388.27
2009年11月
——
——
3097.17
2009年12月
——
——
2938.37
2010年1月
——
1678.61
3469
2010年2月
2677.85
1255.95
2662
To forecast, we substitute the values of BDI, CCBFI and cargo handling capability
into the fitted regression as follows:
Zt  0.0204 X (t 1)  0.0423Y(t 1)  0.2535Y(t 2)  0.5618Z (t 1)  0.0201Z (t 2)
0.0414Z (t 3)  0.0747 Z (t 4)  0.0996Z(t 5)  0.1618Z(t 6)  757.015
From the results, we know that the GZ port cargo handling capability is almost
28,975,500 tons. From statistics of the Guangzhou Port Authority (GPA), we find the
actual value of the GZ cargo was 31.98 million tons in March 2010 and the percentage
error is 9.39%.
From the results, we conclude that there is little relationship among the
Guangzhou cargo port handling capacityt, the international dry bulk shipping freight
rates and coastal dry bulk shipping freight rates. However, there is better relationship
between cargo handling capacity and its own historical data.
Based on the observation of the international shipping market and taking the
analysis of the actual situation in GZ port into consideration, we can conclude that the
dry bulk shipping freight rates have little effects on the GZ cargo port handling capacity.
The possible factors that lead to this conclusion are as follows:
First, in international trade, with cargo ships becoming larger in scale and the rapid
development of information technology, the proportion the transportation costs account
for a much smaller percentage of cargo value. In the United States, transport costs of
shipping account for 4% in 1997 whereas the proportion around 8% in 1974. By 2007
this percentage dropped 100% in three decades (Hummels, 2007).
Second, there is little flexibility in transport costs for the majority of GZ Port's
cargo strategic supplies such as grain, energy and raw materials. The import and export
goods of the four main harbor districts of GZ port are mostly grain, coal, petroleum,
iron ore and other necessities of life and the production or industrial raw materials.
Regardless of the level of transport costs, these import and export goods cannot at
present change radically.
In view of this conclusion, we will remove BDI and CCBFI from our forecasting
model and include the time series of GZ port handling capacity to predict.
Two parameter exponentially weighted moving average (EWMA
forecasting model
EWMA models assume smoothing parameters weight previous observations of
the variable to be predicted (Jarrett, Chapter 2, 1991, and Brockwell and Davis, 2006,)
while two parameter EWMA models will do well when there is a trend in the data
Holt,
).
The simple exponential smoothing is to “smooth out” the irregular fluctuations in
the time series, which is adjusted for the stationary time series, i.e. the time series
without trend, cyclical and seasonal component. Two parameter EWMA models should
be employed when a trend component exists in a time series. Two parameter models
improve forecasting results by adjusting the single EWMA by including and adding the
trend component in the model. The equations for two parameter EWMA forecasting are
given by the following equations.
Ft  Yt  (1   )  Ft 1  Rt 1 
Rt   ( Ft  Ft 1 )  (1   ) Rt 1
Where,
Ft = forecast of the time series for period t  1
Yt  actual value of the time series in period t
Ft -1  forecast of the time series for period t -1
Rt  the best estimate of the trend at time t
  smoothing constant (0    1)
  the trend smoothing factor (0    1)
Although any value of α and β between 0 and 1 are acceptable, however they yield
different results. Often forecasters choose the mean squared error (MSE, Jarrett, 1987,
pp.23-24) and/or the mean absolute deviation (MAD Jarrett, 1987, pp.23-24) as a
measure of forecast accuracy. The smaller the MSE or MAD value is, the better is the
forecast accuracy. Figure 5 below shows the scatterplot of the linear trend in the GZ
port handling capacity.
_________________________________________________________________
Figure 5
Trend chart of the GZ port handling capability from 2004 to 2009
_________________________________________________________________
The GZ port handling capability indicates a linear growth in the port capacity.
Hence, the two parameter EWMA model appears useful in the prediction of the port
capacity. According to the time series combined by 74 sets of data of the GZ port cargo
capacity from January 2004 to February 2010, we do predict an accurate short-term
multi-period ahead model.
_________________________________________________________________
Figure 6
Chart of smoothing values of the GZ cargo port cargo capacity.
Actual value in blue, smoothed value in red (in 10k tons)
_________________________________________________________________
For different values of  and  , the associated and calculated values of MAD
and MSE are shown in Appendix C. If   0.9,   0.1 ; MAD and MSE are at
minimum values, MAD  24.56209 ，and MSE  1067.885 ; or at the optimal
prediction accuracy for the sampled data. If one draws the fits and actual values on a
single plot as shown in Figure 6, the two plots are nearly coincident, indicating that the
prediction model has great accuracy.
On the basis of the two parameter exponential smoothing prediction formula for
period t+i; Fˆt i  Ft  iRt 0    1 ，and according to the data of the actual and
smoothed values, and the trend terms of the GZ port capacity from May 2009 to
February 2010 (Schedule D) respectively, we forecast the GZ port capacity in March
2010 and draw the predicted results based on the data of each month. On this basis, we
calculate the mean values of the predicted results over the past 10 months, 9 months, 8
months ... and 1 month respectively. Finally, we calculate the average of 10 mean
values and make it the final predictions of the GZ port capacity in March 2010.
Calculating the arithmetic mean of the mean values is equivalent to placing different
weights on different months, the closer the months is to predict the greater are the
weights. By this method, the predicting outcomes of the GZ port capacity in March
2010 are shown in Table 4.
According to the above method, we can draw the conclusion that the GZ cargo is
31,719,040 tons in March 2010. From the statistics of GZ Port Authority, we know that
the actual cargo capacity of the GZ port is 31.98 million tons in March 2010 and the
error is 0.816%, achieving very high prediction accuracy.
_________________________________________________________________
Table 4
Partial Data Table of the Guangzhou port cargo throughput in March
2010
Actual
Smoothing
Trend
predicted
Average of predicted
value
value
term
value
May. 2009
3241.72
3237.717
27.47169
3512.434
3406.823
Jun. 2009
3373.94
3363.065
37.25929
3698.398
3395.088
Month
value（Ten
thousand tons）
Jul. 2009
3434.15
3430.767
40.30361
3753.196
3357.175
Aug. 2009
3299.92
3317.035
24.90002
3491.335
3300.6
Sept.2009
3616.29
3588.855
49.59196
3886.406
3268.811
Oct. 2009
3388.27
3413.288
27.07608
3548.668
3145.292
Nov. 2009
3097.17
3131.489
-3.81136
3116.244
3044.448
Dec. 2009
2938.37
2957.301
-20.8491
2894.754
3020.516
Jan. 2010
3469
3415.745
27.08026
3469.906
3083.397
Feb. 2010
2662
2740.083
-43.194
2696.889
2696.889
Mean value
3171.904
_________________________________________________________________
In the same way, we can predict that the Guangzhou port capacity is 32,579,840
tons in April 2010. From the statistics of Guangzhou Port Authority, we know that the
actual value of the Guangzhou cargo is 34.44 million tons in April 2010 and the error of
the prediction is 5.40%.
Tests on the prediction of the GZ port in March 2010 and April 2010, we can
conclude that the two parameter EWMA model has greater prediction accuracy for the
Guangzhou port than the other methods employed before, and ,hence, the best choice
the forecasting models studied for short-term prediction on the samples data.
Comparison between the GZ cargo port capacity and dry bulk freight
In the last section, we make use of three mathematical models in turn to establish
the forecast model for the GZ cargo port. Based on the multi-linear regression
forecasting model and regression and time series forecasting model, taking the values
of CCBFI and BDI reflecting the domestic and international dry bulk freight as
predictors, we obtained the relationship of the two indexes to GZ cargo handling
capacity. In turn we estimated the magnitude of the relations the coastal dry bulk freight
and international shipping dry bulk freight affect the GZ cargo port handling capacity.
In the multiple linear regression model, taking the BDI and CCBFI for last period
as variables for prediction, we study effects which the dry bulk freight for last period
have on the GZ port cargo throughput. The part of the operational report after running
EViews® is shown in Table 5, where X 1t and X 2t denote the values of BDI and
CCBFI for period t respectively and Yt 1 represents the Guangzhou port cargo
throughput for period t +1.
________________________________________________________________
Partial Data Table of the operational report of the multiple regression model
Variable
Coefficient
Std. Error
t-Statistic
Prob.
X 1t
0.036609
0.047248
0.774825
0.4411
X 2t
-0.043952
0.300456
-0.146285
0.8841
If the significance value,   0.05 , the p  value corresponding to the coefficient
of X 1t for the t test is p  value=0. 441> ，suggests that X 1t have on Yt 1 are not
related. Meanwhile the p  value corresponding to the coefficient of X 2t for the t test
is p  value=0. 884> ，also suggests that X 2t and Yt 1 are not related. Therefore, we
may conclude from the multiple linear regression analysis that the international dry
bulk freight and coastal dry bulk freight for last period have little relation to the GZ port
cargo activity.
In the analysis of regression and time series model, we can take the values of BDI,
CCBFI and the GZ port cargo throughput in the past several periods as variables to
predict, and study the effects the domestic and international dry bulk freights in the past
several periods have on the Guangzhou port cargo throughput. We select periods have
the best goodness of fit. The part of the operational report (using EViews® software)
was shown in Table 6.
Where
X (t 1) = value of BDI for period t  1
Y(t 1) = value of CCBFI for period t  1
Y(t  2) = value of CCBFI for period t  2
Z t  value of the Guangzhou port cargo throughput in period t
Variable
Coefficient
Std. Error
t-Statistic
Prob.
X ( t 1)
0.020419
0.028933
0.705738
0.4833
Y( t 1)
0.042297
0.251350
0.168278
0.8670
Y( t 2)
-0.253484
0.220607
-1.149029
0.2554
If   0.05 , the p  value corresponding to the coefficient of X (t 1) for the t test
is p  value=0. 4833> ，which suggests that Y(t 1) and Z t are not related. Similarly,
the p  value corresponding
is
p  value=0. 867>
to
the
，indicating
The p  value corresponding
to
coefficient
that
Y(t 1)
the coefficient
of
Y(t 1)
and
Zt
of
Y(t 2)
for
are
for
the
not
t
test
related.
the t
test
is p  value=0. 2554> ，which suggests that Y(t 2) and Z t are not related. Therefore,
we may conclude that the international dry bulk freight and coastal dry bulk freight in
the past several periods have little effects on the GZ cargo port handling capacity.
Since the coefficient estimates for the multi-linear regression and the time series
regression cannot reject the null hypothesis that they are zero. or the past several
periods that the domestic and international dry bulk freight rates in all have little effects
on the cargo activity of the GZ port and cannot afford a significant role.
Implications and Summary
When making development strategies and policies Guangzhou port, we must give
sufficient consideration to the conclusion that the international and domestic dry bulk
freight rates have little effect on estimating GZ cargo port capacity. Thus, despite the
changing environment in the international shipping market, we see little affect of the
activity in the GZ port development. The GZ port can continue to develop on the basis
of improving its own port facilities. If the port meets the growing demand for cargo
handling capacity, then growth will occur.. Port conditions include waterway facilities,
berthing facilities and harbor management level. Waterway is the lifeline of the port.
The ship channel depth is related directly to whether the ship can reach the harbor. The
port channel of the GZ port is 115 kilometers, which easily get accumulation of silt, we
must attach great importance to deepening sea routes and improving port conditions.
Wharfs, especially the size and number of berths, are important factors in determining
the port handling capacity. GZ Port must pay attention to the construction of new berths
and increasing port handling capacity so as to meet the growing demand for cargo
throughput. Port management level affects not only the utilization efficiency and
economic efficiency of the port, but also affects the level of service and the images of
port companies which cannot be ignored similarly.
Opening up new routes and friendly ports is a means of broaden services scope at
home and increasing supplies. Therefore, it is an integral part of development strategies
of the port. On the one hand, we can open up new routes having development potential
through the cooperation with shipping companies.
For the purpose of forecasting handling capacity, in this study we determined that
good and accurate prediction is warranted for future planning. In turn, the results of our
study indicate that a simple two parameter EWMA model may prove extremely
accurate, easy-to-use and an important model in predicting future needs. Conclusion
At first, in this paper we use 74 sets of data of the Guangzhou cargo port , BDI and
CCBFI from January 2004 to February 2010 to establish the multiple linear regression
model. The analysis on the regression results shows that the BDI and CCBFI have little
effects on the GZ port cargo throughput for next period. According to the preliminary
judgment, there are two reasons for this result. One is the lag on the effects that the dry
bulk shipping freight have on the GZ port cargo throughput and another is the effects of
other factors.
Then, according to the above analysis, we introduce the data of the BDI, CCBFI
and the time series of the GZ cargo port in the past several periods and analyze the
effects the above three kinds of series have on the GZ cargo port by making use of the
regression and time series model. We conclude that dry bulk freight rates in
international shipping market and the Chinese coastal shipping market have little
effects on the GZ cargo port, while the time series is of great help for us to predict the
GZ cargo port . In addition, the prediction accuracy of the model is not large.
In addition, two parameter EWMA requires use of only the historical time series of
the GZ port for te purpose of forecasting. By example tests on the Guangzhou port
cargo in March 2010 and April 2010, we can conclude that the two parameter EWMA
resulted in the best forecasts. Future modeling should consider ARIMA and
Multivariate ARIMA models. The second set of these models may aid in long-term
forecasting. Further, these models should be considered in the light of studies of the
practice of forecasting (De Gooijer and Hyndman, 2006).
References
Anderson, D. R. and Sweeney D.J. (2006) Statistics for Business and Economics, China
Machine Press
Bakshia, G, Panayotov, G and Skoulakis, G (2011) The Baltic Dry Index as a Predictor of
Global Stock Returns, Commodity Returns, and Global Economic Activity, Electronic copy
available at: http://ssrn.com/abstract=1747345
Brockwell, P.J. and Davis, R.A. (2006) Time Series; Theory and Methods, Springer
Chase, R.B. and Jacobs, E. R. (2008) Operations Management for Competitive Advantage
China Machine Press, 477-481.
Chen, N., Zhu, M. and Yu.Z. (2005) “Comparing Two Forecast Models of Ports Handling
Capacity,” Journal of Wuhan University of Technology, 27, 9, 77-79. (In Chinese)
Cheng, L, Wong, Y. (1997) Port Facilities and Container Handling Services: The Hong Kong
Economic Policy Studies, City University of Hong Kong Press
De Gooijer, J.G. and Hyndman, R.J. (2006) “25 years of time series forecasting,” International
Journal of Forecasting, 22, 3, 443-473
Fahrmeir, L., Kneib, T., Lang, S. and Marx, B. (2013) Regression Models, Methods and
Applications, Springer: ISBN 978-3-642-34332-2
Holt, C.C. (1957). "Forecasting Trends and Seasonal by Exponentially Weighted Averages".
Office of Naval Research Memorandum, 52, Reprinted in Hollt, C.C. (January-March, 2004)
“Forecasting Trends and Seasonal by Exponentially Weighted Averages.” International
Journal of Forecasting, 20, 1, 5-10. Doi: 10.1016/j.ijforecast.2003.09.015
Hummels, D. (2007) Transportation Costs and International Trade in the Second Era of
Globalization Journal of Economic Perspectives，135-136
Hui, E., Seabrooke, W. and Woong, K. (2004) “Forecasting Cargo Throughput for the Port of
Hong Kong: Error Correction Model Approach,” Journal of Urban Planning and
Development, 130,4, 195-203
J. Jarrett (1991) Business Forecasting Methods, Basil Blackwell; Oxford
Lam, L. Pan, et al. (2004) Forecasts and Reliability Analysis of Port Cargo Throughput in Hong
Kong. Journal of Urban Planning and Development, 130, 3, l33-144
Pindyck, R.S. and Rubinfeld, D.L. (2010).Econometric Model and Economic Forecasts, 4
ed., McGraw-Hill: ISBN 0070502080 | 0-07-050208-0
th
Sun, L. (2010) “The research on a double forecasting model of port cargo throughput,” World
Journal of Modeling and Forecasting, 6, 1, 57-62
Sun, L. and Wang, Z. (2007) Studying the Probability Distribution of the port handling
capacity in the port. Proceedings of SSMSSD, 6, 1869-1873. (In Chinese)
Sun, Y. G. Zheng, G. (2007) “Research on Forecast of Port Throughput Capacity Based on
Grey Theory,” China Water Transport, 5, 4, 160-162.
Wang.C. (2006) “Port Handling Capacity Forecasting Based on Unascertained Number
Regressive Model Communication Standardization”
Xu, C., Y. Yan, Y. and Zhang, P.(2006) Prediction Model of Port Throughput Based on
System Dynamics. Port & Waterway Engineering, 32, 4, 26-29. (In Chinese)
Yuan, E., and Y. Xie, Y. (2007) On Ports Throughput Probability Forecasting Model Port
and Waterway Engineering, 33, 5, 1028-1031 (In Chinese)
Yang, S. Guan, K (2005) “Application of Spreadsheet to Forecasting Port Throughput,” Wuhan
Journal of University of Technology (Transportation Science and Engineering)
Yuan, E., and Y. Xie, Y. (2007) On Ports Throughput Probability Forecasting Model Port
and Waterway Engineering, 33, 5, 1028-1031 (In Chinese)
Appendix A
month
1/2004 to 2/2010, BDI、CCBFI and Guangzhou Port Volume
BDI
CCBFI volume（10k ton）
month
BDI
CCBFI volume（10k ton）
2004 年 1 月 5227.48 1420.2
1479.49
2007 年 2 月
4383.00 1799.58
2888.06
2004 年 2 月 5450.05 1514.27
1680.8
2007 年 3 月
5168.30
1745.4
2883.76
2004 年 3 月 5145.27 1527.66
1728.68
2007 年 4 月
5800.82 1696.78
3009.24
2004 年 4 月 4493.30 1621.58
2024.6
2007 年 5 月
6425.22 1710.18
3117.29
2004 年 5 月 3595.58 1717.61
1828.14
2007 年 6 月
5761.35 1781.62
2646
2004 年 6 月 2916.26 1668.26
1771.67
2007 年 7 月
6554.65 1871.19
2909.42
2004 年 7 月 3773.73 1609.65
1741.74
2007 年 8 月
7207.43 1951.33
3167.52
2004 年 8 月 4169.00 1612.17
1953.18
2007 年 9 月
8574.05 2105.02
3203.01
2004 年 9 月 4141.62 1592.84
2294.16
2007 年 10 月 10668.24 2161.09
3172.96
2004 年 10 月 4539.43 1597.48
2417.48
2007 年 11 月 10550.86 2281.12
3174.11
2004 年 11 月 5302.57 1527.39
2394.02
2007 年 12 月 9842.93 2498.87
3310.98
2004 年 12 月 5571.24 1514.8
2182.26
2008 年 1 月
7169.82 2607.72
3438.57
2005 年 1 月 4501.90 1549.99
1941.35
2008 年 2 月
7020.53 2544.72
2652.85
2005 年 2 月 4620.20 1638.13
1388.23
2008 年 3 月
8086.56 2331.69
3276.28
2005 年 3 月 4676.95 1657.83
1890.63
2008 年 4 月
8301.00 2237.68
3222.9
2005 年 4 月 4563.21 1657.94
2043.36
2008 年 5 月 10909.28 2550.81
3357.47
2005 年 5 月 3698.95 1639.45
1956.9
2008 年 6 月 10176.75 2743.37
3141.04
2005 年 6 月 2746.09 1581.3
2038.24
2008 年 7 月
8935.74 2144.28
2971.95
2005 年 7 月 2220.10 1462.45
2045.13
2008 年 8 月
7402.50 1994.97
2396.23
2005 年 8 月 2203.86 1437.4
2059.21
2008 年 9 月
5154.26 1737.72
2485.39
2005 年 9 月 2803.55 1487.95
2683.88
2008 年 10 月 1694.52 1491.43
3234.19
2005 年 10 月 3164.60 1508.51
2708.86
2008 年 11 月
818.95
1212.09
3397.22
2005 年 11 月 2916.82 1446.46
2837.56
2008 年 12 月
742.41
1196.74
3325.76
2005 年 12 月 2624.00 1430.58
3011.59
2009 年 1 月
904.52
1672.6
2700.8
2006 年 1 月 2263.85 1474.66
2756.11
2009 年 2 月
1815.85 1097.03
2397.25
2006 年 2 月 2435.71 1459.56
2449.89
2009 年 3 月
1957.82 1109.64
2851.81
2006 年 3 月 2595.27 1474.24
2409.02
2009 年 4 月
1644.15 1164.36
3221.03
2006 年 4 月 2472.53
2534.29
2009 年 5 月
2539.84 1057.34
3241.72
2006 年 5 月 2444.58 1541.14
2890.31
2009 年 6 月
3822.55 1009.39
3373.94
2006 年 6 月 2705.69 1533.52
2826.34
2009 年 7 月
3361.87 1039.82
3434.15
2006 年 7 月 3051.46 1533.44
2267.59
2009 年 8 月
2672.19 1146.08
3299.92
2006 年 8 月 3692.64 1530.77
2994.18
2009 年 9 月
2357.67 1182.44
3616.29
2006 年 9 月 4045.75 1541.2
2,662
1505
2009 年 10 月 2847.94
1136.5
3388.27
2006 年 10 月 4031.67 1575.7
2845.27
2009 年 11 月 3940.57 1426.14
3097.17
2006 年 11 月 4183.38 1636.95
2923.7
2009 年 12 月 3572.39 2110.29
2938.37
2006 年 12 月 4332.00 1750.04
3058.98
2010 年 1 月
3168.35 1678.61
3469
2007 年 1 月 4488.47 1816.27
3115.68
2010 年 2 月
2677.85 1255.95
2662
sources：Guangzhou Port Administration, Baltic Shipping Exchange, Shanghai Shipping
Exchange
____________________________________________________________________________________
Appendix B
n
4
5
6
7
8
m
j
1
2
R 2 for j, m, n in the model
3
4
5
6
7
8
1
0.574322 0.57507 0.576359 0.577417 0.570042 0.568904 0.582043 0.574016
2
0.571862 0.56948 0.573204 0.576881 0.569641 0.570195 0.58272 0.574547
3
0.56865 0.574063 0.571602 0.573354 0.565819 0.566573 0.575668 0.567195
4
0.561618 0.567321 0.563693 0.566406 0.558864 0.558976 0.567643 0.558917
5
0.553998 0.560441 0.555967 0.558638 0.551321 0.550555 0.558839 0.549745
6
0.546671 0.552424 0.548229 0.552015 0.546939 0.543328 0.549877 0.540372
7
0.538644 0.54596 0.540833 0.543507 0.537846 0.536698 0.550158 0.541292
1
0.580762 0.585571 0.585364 0.583872 0.577231 0.57353 0.584602 0.576648
2
0.579208 0.580031 0.581978 0.582603 0.576245 0.573858 0.58444 0.576298
3
0.572407 0.57788 0.576062 0.575908 0.56913 0.566937 0.576157 0.567679
4
0.566713 0.572726 0.568302 0.567883 0.560947 0.558434 0.567521 0.558726
5
0.558781 0.565266 0.560209 0.560028 0.55358 0.550096 0.558619 0.549437
6
0.552316 0.55776 0.553202 0.554133 0.551629 0.544924
7
0.544683 0.552062 0.54646 0.546002 0.542712 0.537775 0.549757 0.540786
1
0.585459 0.587775 0.585246 0.582677 0.575153 0.570816 0.580637 0.572393
2
0.584017 0.582589 0.581899 0.581286 0.573861 0.570752 0.580124 0.571681
3
0.576853 0.578676 0.575294 0.574047 0.566261 0.56327 0.571558 0.562761
4
0.569315 0.571472 0.567166 0.566705 0.558762 0.555245 0.563001 0.553992
5
0.562085 0.564795 0.559549 0.558123 0.550111 0.546052 0.553728 0.544329
6
0.556044 0.557549 0.552948 0.552649 0.547548 0.54041 0.545239 0.53534
7
0.547807 0.550604 0.545085 0.543775 0.538118 0.532224 0.542766 0.533483
1
0.578604 0.580281 0.577617 0.574825 0.567017 0.562423 0.572258 0.563678
2
0.576876 0.574867 0.57405 0.573303 0.565604 0.562171 0.571584 0.562792
3
0.569522 0.570729 0.567161 0.565766 0.557686 0.554364 0.562662 0.553492
4
0.561772 0.563241 0.558724 0.558073 0.549811 0.546001 0.55372 0.544315
5
0.554848 0.556477 0.550746 0.549117 0.540743 0.536446 0.544058 0.534241
6
0.549191 0.549182 0.543925 0.543504 0.538169 0.530459 0.535298 0.524952
7
0.54083 0.542012 0.535768 0.534286 0.52836 0.521855 0.532509 0.522776
1
0.571682 0.574314 0.570559 0.568005 0.559719 0.555164 0.567488 0.558574
2
0.570152 0.568805 0.566737 0.566337 0.558076 0.554733 0.566611 0.557475
3
0.56242 0.564322 0.559653 0.558577 0.54992 0.546728 0.557394 0.547859
4
0.554286 0.55639 0.550891 0.550598 0.541704 0.538001 0.548086 0.538253
0.5505
0.540857
5
0.54679 0.548931 0.54242
0.54143 0.532417 0.528162 0.538046 0.527764
6
0.542318 0.542638 0.536508 0.537981 0.532557 0.524211 0.530817 0.519983
7
0.534152 0.536238 0.52888 0.529136 0.522934 0.516449 0.531704 0.521489
Appendix C 2004 to 2010 Exponential Smoothing for GZ Port Volume
No. α β
MAD
MSE
No. α β
MAD
MSE
No. α β
MAD
MSE
1
0.1 0.1 230.9088 88453.96
31 0.4 0.1 150.1611 37302.69
61 0.7 0.1 74.34868 9307.38
2
0.1 0.2 240.6077 91597.43
32 0.4 0.2 159.1768 40413.11
62 0.7 0.2 76.97842 10052.84
3
0.1 0.3 248.5674 98460.8
33 0.4 0.3 166.4069
63 0.7 0.3 79.73506 10792.74
4
0.1 0.4 256.5708 105538.4
34 0.4 0.4 170.6041 45620.81
64 0.7 0.4 82.54973 11552.27
5
0.1 0.5 265.7206 115184.8
35 0.4 0.5 173.1541 47944.29
65 0.7 0.5 85.40904 12339.17
6
0.1 0.6 277.5096 128283.8
36 0.4 0.6 176.745 50318.43
66 0.7 0.6 88.33114 13140.98
7
0.1 0.7 294.4708 142693.1
37 0.4 0.7 181.624 52915.18
67 0.7 0.7 91.28124 13934.59
8
0.1 0.8 308.7417
38 0.4 0.8 187.0292 55884.67
68 0.7 0.8 93.72225 14696.96
9
0.1 0.9 314.7122 162027.5
39 0.4 0.9 193.1834 59318.76
69 0.7 0.9 95.35599 15413.55
314.9717 164869.3
40 0.4 1 199.4062 63233.41
70 0.7 1 97.04101 16082.17
11 0.2 0.1 204.5025 67906.86
41 0.5 0.1 125.2613 25732.41
71 0.8 0.1 49.30028 4185.838
12 0.2 0.2 211.5926 74229.98
42 0.5 0.2 132.4981 27785.12
72 0.8 0.2 51.04407 4525.441
13 0.2 0.3 220.6944 81635.52
43 0.5 0.3 136.711 29685.58
73 0.8 0.3 52.78538 4865.94
14 0.2 0.4 230.1412 88592.25
44 0.5 0.4 139.7643 31536.46
74 0.8 0.4 54.51865 5213.739
15 0.2 0.5 239.7914 94051.16
45 0.5 0.5 144.1503 33469.88
75 0.8 0.5 56.4979 5567.977
16 0.2 0.6 249.5269 98713.92
46 0.5 0.6 148.8134 35579.93
76 0.8 0.6 58.35731 5922.854
17 0.2 0.7 260.1872 103304.2
47 0.5 0.7 154.3846 37911.97
77 0.8 0.7 59.83785 6272.718
18 0.2 0.8 268.7438
107465
48 0.5 0.8 159.2187 40453.47
78 0.8 0.8 61.25162 6615.581
19 0.2 0.9 271.8719
110736
49 0.5 0.9 164.3803 43138.76
79 0.8 0.9 62.78413 6954.31
271.8044 113138.5
50 0.5 1 169.5226 45865.95
80 0.8 1 64.02011 7295.902
21 0.3 0.1 174.9438 51382.12
51 0.6 0.1 99.93618 16459.55
81 0.9 0.1 24.56209 1067.885
22 0.3 0.2 185.3171 56104.21
52 0.6 0.2 104.512 17767.23
82 0.9 0.2 25.36117 1156.758
23 0.3 0.3 195.3688 60483.33
53 0.6 0.3 107.408 19034.65
83 0.9 0.3 26.2977 1246.816
24 0.3 0.4 203.1649 64111.89
54 0.6 0.4 111.2502 20327.9
84 0.9 0.4 27.23918 1339.241
25 0.3 0.5 208.4073 67211.68
55 0.6 0.5 115.5866 21697.36
85 0.9 0.5 28.18603 1434.007
26 0.3 0.6 211.0237 69917.51
56 0.6 0.6 119.0632 23152.82
86 0.9 0.6 29.16206 1530.792
27 0.3 0.7 211.9083 72325.71
57 0.6 0.7 122.5493 24667.26
87 0.9 0.7 30.02898 1629.985
28 0.3 0.8 213.2795 74605.11
58 0.6 0.8 126.8816 26188.48
88 0.9 0.8 30.81781 1733.077
29 0.3 0.9 217.3025 76981.97
59 0.6 0.9 130.1989 27656.08
89 0.9 0.9 31.50343 1842.601
30 0.3 1
60 0.6 1 133.3653 29018.01
90 0.9 1 32.29509 1961.894
10 0.1 1
20 0.2 1
154803
223.2911 79721.59
43174
Appendix D 1/2004 to 2/2010 Exponential Smoothing with Trend for
GZ Port Volume
month
Actual value
smoothing
Trend item
Absolute error
variance
2004 年 1 月
1479.49
1479.49
2004 年 2 月
1680.80
1660.669
58.44
20.131
405.257161
2004 年 3 月
1728.68
1727.7225
59.29775
0.9575
0.91680625
2004 年 4 月
2024.60
2000.842
80.6799275
23.757975
564.4413761
2004 年 5 月
1828.14
1853.4782
57.87555178
25.3381952
642.0241385
2004 年 6 月
1771.67
1785.6384
45.30401454
13.9683747
195.1154918
2004 年 7 月
1741.74
1750.6602
37.27579951
8.92023892
79.57066247
2004 年 8 月
1953.18
1936.6556
52.14775605
16.5243962
273.0556683
2004 年 9 月
2294.16
2263.6243
79.62985366
30.535664
932.4267766
2004 年 10 月
2417.48
2410.0574
86.31017659
7.42258103
55.09470922
2004 年 11 月
2394.02
2404.2548
77.09889299
10.2347596
104.7503032
2004 年 12 月
2182.26
2212.1694
50.18046426
29.9093653
894.5701299
2005 年 1 月
1941.35
1973.45
21.29047961
32.099983
1030.408906
2005 年 2 月
1388.23
1448.881
-33.29546202
60.6510463
3678.549412
2005 年 3 月
1890.63
1843.1256
9.458535395
47.5044416
2256.67197
2005 年 4 月
2043.36
2024.2824
26.62836695
19.0775906
363.9544638
2005 年 5 月
1956.90
1966.3011
18.16739708
9.40107763
88.38026067
2005 年 6 月
2038.24
2032.8628
23.00683436
5.37715253
28.91376931
2005 年 7 月
2045.13
2046.204
22.04026299
1.07396818
1.153407658
2005 年 8 月
2059.21
2060.1134
21.22718219
0.90342312
0.816173329
2005 年 9 月
2683.88
2623.6261
75.45572771
60.2539395
3630.537222
2005 年 10 月
2708.86
2707.8822
76.33576677
0.97782118
0.956134252
2005 年 11 月
2837.56
2832.2258
81.13655166
5.33420544
28.45374768
2005 年 12 月
3011.59
3001.7672
89.9770405
9.82276538
96.48671966
2006 年 1 月
2756.11
2789.6734
59.76995574
33.5634275
1126.503666
2006 年 2 月
2449.89
2489.8453
23.81015125
39.9553383
1596.429061
2006 年 3 月
2409.02
2419.4835
14.39295719
10.463549
109.4858568
2006 年 4 月
2534.29
2524.2487
23.43017164
10.0413494
100.8286975
2006 年 5 月
2890.31
2856.0469
54.26697763
34.2631178
1173.96124
2006 年 6 月
2826.34
2834.7374
46.70933025
8.39738599
70.51609139
2006 年 7 月
2267.59
2328.9757
-8.537774215
61.3856716
3768.200681
2006 年 8 月
2994.18
2926.8058
52.09901502
67.3742103
4539.284208
2006 年 9 月
2662.00
2693.6905
23.57758259
31.6904805
1004.286553
2006 年 10 月
2845.27
2832.4698
35.09775691
12.8001937
163.8449586
2006 年 11 月
2923.70
2918.0868
40.14967622
5.61324368
31.50850459
2006 年 12 月
3058.98
3048.9056
49.21659729
10.0743567
101.4926638
To be continued
D continued
month
Actual value
smoothing
Trend item
Absolute error
variance
2007 年 1 月
3115.68
3113.9242
50.79679565
1.75577595
3.082749169
2007 年 2 月
2888.06
2915.7261
25.89730387
27.666102
765.4131982
2007 年 3 月
2883.76
2889.5463
20.68959735
5.78634058
33.48173736
2007 年 4 月
3009.24
2999.3396
29.59996293
9.90040621
98.01804306
2007 年 5 月
3117.29
3108.455
37.55150283
8.83504433
78.05800827
2007 年 6 月
2646.00
2696.0006
-7.449078438
50.0006459
2500.064585
2007 年 7 月
2909.42
2887.3332
12.4290805
22.0868433
487.8286451
2007 年 8 月
3167.52
3140.7442
36.52727914
26.7757763
716.9421952
2007 年 9 月
3203.01
3200.4362
38.84374389
2.57384971
6.624702346
2007 年 10 月
3172.96
3179.592
32.87495341
6.63198942
43.98328363
2007 年 11 月
3174.11
3177.9457
29.42282856
3.83569428
14.71255063
2007 年 12 月
3310.98
3300.6189
38.7478615
10.3611477
107.353382
2008 年 1 月
3438.57
3428.6497
47.67615726
9.92032862
98.41291996
2008 年 2 月
2652.85
2735.1976
-26.43666732
82.3475829
6781.124404
2008 年 3 月
3276.28
3219.5281
24.64005028
56.7519084
3220.779112
2008 年 4 月
3222.90
3225.0268
22.72591752
2.12681418
4.523338572
2008 年 5 月
3357.47
3346.4983
32.60047166
10.9717268
120.3787896
2008 年 6 月
3141.04
3164.8459
11.17518463
23.8058745
566.7196599
2008 年 7 月
2971.95
2992.3571
-7.191210691
20.4071059
416.4499717
2008 年 8 月
2396.23
2455.1236
-60.19544126
58.8935895
3468.454887
2008 年 9 月
2485.39
2476.3438
-52.0538746
9.04618517
81.8334662
2008 年 10 月
3234.19
3153.2
20.83713078
80.990006
6559.381068
2008 年 11 月
3397.22
3374.9017
40.92358954
22.3182875
498.1059578
2008 年 12 月
3325.76
3334.7665
32.81771236
9.0065302
81.11758629
2009 年 1 月
2700.80
2767.4784
-27.19286947
66.6784243
4446.012261
2009 年 2 月
2397.25
2431.5536
-58.0660694
34.3035555
1176.733918
2009 年 3 月
2851.81
2803.9777
-15.01704315
47.8322514
2287.924273
2009 年 4 月
3221.03
3177.8231
23.86919336
43.2069295
1866.838753
2009 年 5 月
3241.72
3237.7172
27.47168961
4.00277361
16.02219657
2009 年 6 月
3373.94
3363.0649
37.25928717
10.8751084
118.2679827
2009 年 7 月
3434.15
3430.7674
40.30361108
3.38258212
11.44186182
2009 年 8 月
3299.92
3317.0351
24.90001847
17.1151029
292.9267471
2009 年 9 月
3616.29
3588.8545
49.59195755
27.4354879
752.7059943
2009 年 10 月
3388.27
3413.2876
27.07607528
25.017647
625.8826599
2009 年 11 月
3097.17
3131.4894
-3.811359723
34.3193722
1177.81931
2009 年 12 月
2938.37
2957.3008
-20.84908085
18.9308013
358.375236
2010 年 1 月
3469.00
3415.7452
27.08026432
53.254828
2836.076701
2010 年 2 月
2662.00
2740.0825
-43.19402496
78.0825436
6096.883621
Founded in 1892, the University of Rhode Island is one of eight land, urban, and sea grant
universities in the United States. The 1,200-acre rural campus is less
than ten miles from Narragansett Bay and highlights its traditions of
natural resource, marine and urban related research. There are over
14,000 undergraduate and graduate students enrolled in seven degreegranting colleges representing 48 states and the District of Columbia.
More than 500 international students represent 59 different countries.
Eighteen percent of the freshman class graduated in the top ten percent
of their high school classes. The teaching and research faculty numbers
over 600 and the University offers 101 undergraduate programs and 86
advanced degree programs. URI students have received Rhodes,
Fulbright, Truman, Goldwater, and Udall scholarships. There are over 80,000 active alumnae.
The University of Rhode Island started to offer undergraduate business
administration courses in 1923. In 1962, the MBA program was introduced and the PhD
program began in the mid 1980s. The College of Business Administration is accredited by
The AACSB International - The Association to Advance Collegiate Schools of Business in
1969. The College of Business enrolls over 1400 undergraduate students and more than 300
graduate students.
Mission
Our responsibility is to provide strong academic programs that instill excellence,
confidence and strong leadership skills in our graduates. Our aim is to (1)
promote critical and independent thinking, (2) foster personal responsibility and
(3) develop students whose performance and commitment mark them as leaders
contributing to the business community and society. The College will serve as a
center for business scholarship, creative research and outreach activities to the
citizens and institutions of the State of Rhode Island as well as the regional,
national and international communities.
The creation of this working paper series
has been funded by an endowment
established by William A. Orme, URI
College of Business Administration,
Class of 1949 and former head of the
General Electric Foundation. This working
paper series is intended to permit faculty
members to obtain feedback on research
activities before the research is submitted to
academic and professional journals and
professional associations for presentations.
An award is presented annually for the most
outstanding paper submitted.
Ballentine Hall
Quadrangle
Univ. of Rhode Island
Kingston, Rhode Island