Professional Documents
Culture Documents
PAGE
1. Learning Issues
1.1 What is Forecasting 3
1.2 Techniques in Forecasting 3
1.3 Time Series Technique 4
1.4 Econometric Technique 6
1.5 Forecast variable using time series & econometric technique 7
1.6 Evaluation of Forecast Accuracy 10
2. Introduction 12
2
7. References 20
8. Appendix
1. Learning Issues
To overcome forecasting problems caused by error and bias, a variety of forecast methods can
be employed. Some forecast techniques are basically quantitative, and some are qualitative.
The best forecast method for any particular task depends on the nature of the forecasting
problem. Several factors that we need to consider such as distance in future that one must
forecast, the lead time available for making decisions, the level of accuracy required, the
quality of data available for analysis, and the cost and benefits associated with the forecasting
problem.
The panel consensus method assumes that several experts can arrive at forecasts that are
superior to those that individuals generate.
Although the panel consensus method often results in forecasts that embody the collective
wisdom of consulted experts, it can be unfavorably affected by the forceful personality of one
or more key individuals. To overcome, Delphi method has been developed. It is a method
that uses forecasts derived from an independent analysis of expert opinion. . In this method,
members of a panel of experts individually receive a series of questions relating to the
underlying forecasting problem. Responses are analyzed by an independent party, who then
3
tries to elicit a consensus opinion by providing feedback to panel members in a manner that
prevents direct identification of individual positions. This method helps limit the steamroller
or bandwagon problems of the basic panel consensus approach.
On the other hand, Quantitative analyses are base on measurable and verifiable data such as
earnings, revenue, wages, market share, etc. to be use for forecasting. Therefore it is much
more objective than qualitative analysis. There are mainly 2 methods that are commonly use:
time series methods and econometric methods. Time series methods generally use data from
the past and usually a periodic data to predict future events. For example, assume that you are
interested in knowing how long a recession will last. You might look at all past recessions and
the events leading up to and surrounding them and then, from that data, try to predict how
long the current recession will last.
The other quantitative method which is econometric methods statistically identify the
relationships between variables and how it can affect the other one or more variables when it
changes. An econometric model is a way of determining the strength and statistical
significance of a hypothesized relationship. These models are used extensively in economics
to prove, disprove, or validate the existence of a casual relationship between two or more
variables. It is obvious that this model is highly mathematical, using different statistical
equations. There are also called regression models.
What makes this different from qualitative analysis is researcher knows clearly in advance
what they are looking for, and data is in the form of numbers and statistics therefore its data is
more efficient, and able to test hypotheses.
Times series analysis is forecasting future events using a series of observations taken on an
economic variable at a various past points in time. Observations may be taken every hour,
day, week, month or year or at any other regular or irregular interval. Time series analyses are
adequate forecasting tools if demand has shown a consistent pattern in the past that is
expected to recur in the future. For example, by using a regular time interval (monthly
4
observations from Jan 2000 till July 2010), we can forecast the production of palm oil from
August 2010 till December 2010 by using time-series analysis technique.
In the analysis of time-series data, a graph consist of time (on the horizontal axis) versus the
values of the dependent variable (on the vertical axis) is plot. This graph can be decomposed
into 4 components that are: - Secular trends, cyclical variations, seasonal effects, and irregular
component (Random fluctuations) and is represent by the formula below.
i) Secular trends
It is the long run changes in an economic time-series. A trend can be positive or negative
depending on the time series exhibits an increasing long term pattern (uptrend) or a
decreasing long term patter (downtrend). A trend can categorize to 3 types that is
a) Linear trend, b) constant rate of growth trend and c) Declining rate of growth trend.
a) Linear trends
Linear trends are mostly estimated by using least-squares regression analysis to provide an
equation of a straight line of “best fit”. The equation of a linear time trend is given in the
general form
Ŷ t = α+ β t
Ŷ t is the forecast or predicted value for period t, Α is the y intercept or constant term, t is the
unit of time, and β is an estimation of the trend factor,
To estimate the parameters directly with the ordinary least squares method, taking logarithms
of both sides of the equation gives
log Ŷ = log Ŷ0 + t . log (1+g)
where Ŷ t = α+ β t , whose parameters can be estimated using standard linear
regression techniques.
5
c) Declining rate of growth trends is represented by the equation
Ŷ=e β1- β2/t
Using linear regression techniques, one can specify a semilog estimation equation
logŶt = β1- β2(1/t)
ii) Seasonality
Seasonality means a particular pattern of fluctuations (in a year) that are more or less
consistent from the past of years. Seasonal components capture the regular pattern of
variability in the time series within one-year periods. For example, an drastic increase of
production of palm oil in the month of January is repeating itself in every January in every
other past years. Seasonal variation can estimate by using ratio to trend method and dummy
variables.
Ratio to trend method assumes that the trend value is multiplied by the seasonal effect or
seasonal index.
A dummy variable is a variable that normally takes on one of two values – either 0 or 1. This
method assumes that the seasonal effects are added to the trend value. If a time series consists
of quarterly data, then the following model could be used to adjust for seasonal effects:
Irregular components are any fluctuations not classified as one of the above and are a
component in time series that is unexplainable therefore it is unpredictable. Estimation of this
component can be expected only when its variance is not too large. Otherwise, it is not
possible to decompose the series.
6
1.4 Econometric Technique
Econometric techniques combine economic theory with statistical tools to predict economic
relationships. Econometric models may vary in their level of sophistication from simple to the
extremely complex. Econometric techniques express relevant economic relations (between
the variables) in the form of an equation. The relationships between the variables can be
determined using regression methods.
An advantage of econometric models is that they predict not only the direction of change in
an economic series, but also the magnitude of that change. How each individual variable
affect each other are due to the cause and effect relations. This means the forecaster must deal
with causal relations. This produces logical consistency in the forecast model and increases
reliability.
In Time-Series Models we presume to know nothing about the causality that affects the
variable we are trying to forecast. Instead, we examine the past behavior of a time series in
order to infer something about its future behavior. The method used to produce a forecast may
involve the use of a simple deterministic model such as a linear extrapolation or the use of a
complex stochastic model for adaptive forecasting. The decision to build a time-series model
usually occurs when little or nothing is known about the determinants of the variable
being studied, when a large number of data points are available (usually over 50), and
when the model is to be used largely for short-term forecasting.
Econometric models are sets of simultaneous regressions models. The model is formulated
so that its parameters can be estimated if one makes the assumption that the model is
correct.Econometrics is mostly studying the issue of causality where one identify a causal
relation between an outcome and a set of factors that determined this outcome.
Scatter Diagram: A graphical representation of the pairs of data called a scatter diagram can be
drawn to gain an overall view of the problem. Is there an apparent relationship? Direct? Inverse? If
the points lie within a band described by parallel lines, we can say there is a linear relationship
between the pair of x and y values. If the rate of change is generally not constant, then the
relationship is curvilinear.
Plot the data based on data collected from 2000 to 2009 as we need to test the forecast for the year
2010.The plotted data showed:
8
Least-Squares Method:
To predict the mean y-value for a given x-value, we need a line which passes through the mean
value of both x and y and which minimizes the sum of the distance between each of the points and
the predictive line. Such an approach should result in a line which we can call a "best fit" to the
sample data.
Well-fitting regression model results in predicted values close to the observed data values. Three
statistics are used in Ordinary Least Squares (OLS) regression to evaluate model fit: R-squared,
the overall F-test, and the Root Mean Square Error (RMSE).
a) R squared is taken as a measure of the “ goodness of fit” of the regression. The closer to 1 is
R squared, the better the fit.
9
Ŷ = 850.7 +5.753t
The F-test
The F-test evaluates the null hypothesis that all regression coefficients are equal to zero versus the
alternative that at least one does not. An equivalent null hypothesis is that R-squared equals zero.
A significant F-test indicates that the observed R-squared is reliable, and is not a spurious result of
oddities in the data set. Thus, the F-test determines whether the proposed relationship between the
response variable and the set of predictors is statistically reliable, and can be useful when the
research objective is either prediction or explanation.
F= 192.585125126631
Significance of F=1.4758678255093Ex10-26
Most important to evaluate the forecast error by using the appropriate statistical tests. We define the best
forecast as the one which yields the forecast error with the minimum variance. In the single-equation
regression model, ordinary lest-squares estimation yields the best forecast among all linear unbiased
estimators having minimum mean-square error.
To determine the accuracy or reliability of a forecasting model is to see the magnitude of the differences
between observed actual(At) and the forecasted values (Ft )
Forecast Error :
10
• et = At – Ft
For this forecast in production of palm oil, we used the RMSE is the square root of the
variance of the residuals. It indicates the absolute fit of the model to the data–how close the
observed data points are to the model’s predicted values. Whereas R-squared is a relative measure
of fit, RMSE is an absolute measure of fit. As the square root of a variance, RMSE can be
interpreted as the standard deviation of the unexplained variance, and has the useful property of
being in the same units as the response variable. Lower values of RMSE indicate better fit. RMSE
is a good measure of how accurately the model predicts the response, and is the most important
criterion for fit if the main purpose of the model is prediction.
11
Given that:
Alternatively, one can directly compute the MAPE or RMSPE over the validation data and get
the percentage that way. In both one has to be sure that the respective measures are multiplied
by 100 in order to convert them to percentages.
2 Introduction
Forecasting is an integral part of economic decision making especially in planning both long
terms and short terms. Forecasts may be used in numerous ways; specifically, individuals may
use forecasts to try to earn income from speculative activities, to determine optimal government
policies, or to make business decisions.
12
3.1 Graph Plotting and time series component (seasonal,trend,cycle,irregular )
13
ii.Forecast of monthly production of crude palm oil for September 2010 ( t=129 ) refer
Appendix 1
ii.Forecast of monthly production of crude palm oil for October 2010 ( t=130 ) refer
Appendix 1
ii.Forecast of monthly production of crude palm oil for November 2010 ( t=131 ) refer
Appendix 1
ii.Forecast of monthly production of crude palm oil for December 2010 ( t=132 ) refer
Appendix 1
These data indicate that , on the average , August crude palm oil production have been 10.3%
higher than the trend value. The Seasonal Adjustment Factor is 1.103
14
ii.Seasonal Adjustment Factor for month of September
September Forecast Actual Actual / Forecast
2000 918.816 1,110.724 1.2089
2001 981.792 1,100.735 1.1211
2002 1,044.768 1,240.960 1.1878
2003 1,107.744 1,297.033 1.1709
2004 1,170.720 1,488.215 1.2712
2005 1,233.696 1,431.655 1.1605
2006 1,296.672 1,610.040 1.2417
2007 1,359.648 1,602.065 1.1783
2008 1,422.624 1,579.442 1.1102
2009 1,485.600 1,557.764 1.0486
2010 1,548.572
SUM 11.6991
SUM AVRG 1.170
These data indicate that , on the average , September crude palm oil production have been
17.0% higher than the trend value. The Seasonal Adjustment Factor is 1.170
These data indicate that , on the average , October crude palm oil production have been
16.8% higher than the trend value. The Seasonal Adjustment Factor is 1.168
These data indicate that , on the average , November crude palm oil production have been
9.4% higher than the trend value. The Seasonal Adjustment Factor is 1.094
These data indicate that , on the average , December crude palm oil production have been
(1-0.973 ) = 2.7% lower than the trend value. The Seasonal Adjustment Factor is 0.973
i.Forecast of monthly production of crude palm oil for August 2010 are seasonally adjusted
upward by 10.3% to :
1,543.324 ( ' 000 tonnes) x 1.103 = 1,702.286( ' 000 tonnes) monthly Production of Crude
palm oil
ii.Forecast of monthly production of crude palm oil for September 2010 are seasonally
adjusted upward by 17.0% to :
1,548.572 ( ' 000 tonnes) x 1.170 = 1,811.829( ' 000 tonnes) monthly Production of Crude
palm oil
16
ii.Forecast of monthly production of crude palm oil for October 2010 are seasonally
adjusted upward by 16.8%
1,553.820 ( ' 000 tonnes) x 1.168 = 1,814.862 ( ' 000 tonnes) monthly Production of Crude
palm oil
ii.Forecast of monthly production of crude palm oil for November 2010 are seasonally
adjusted upward by 9.4% to :
1,559.068 ( ' 000 tonnes) x 1.094=1,705.620 ( ' 000 tonnes) monthly Production of Crude
palm oil
ii.Forecast of monthly production of palm oil for December 2010 are seasonally adjusted
downward by 2.7% to :
1,564.316 ( ' 000 tonnes) x 0.973=1,522.079 ( ' 000 tonnes) monthly Production of Crude
palm oil
Where
PBF= Retail Price of Beef (RM/Kg)
PMT= Retail Price of Mutton (RM/Kg)
INC= Per Capita Income (RM)
QMT= Per Capita Consumption of Mutton (Kg)
17
Analysis in Meat Consumption
P-value = 0.1112 which means that this variable is more than 5% significant level
therefore we accept Ho. We conclude no relationship exists between Per capita
consumption of mutton and retail price of mutton.
P-value = 0.2545 which means that this variable is more than 5% significant level
therefore we accept Ho. We conclude no relationship exists between per capita
consumption of mutton and retail price of beef.
P-value = 3.809E-5 which means that this variable is less than 5% significant level
therefore we reject Ho. We conclude that at the 1 percent level of significant a linear,
positive relationship exists between Per capita consumption of mutton and retail price of
mutton.
Null hypothesis Ho: All β = 0 (no relationship between all variable and QMT)
Alternative hypothesis Ha: At least one β ≠ 0 (at least one relationship between variable
and QMT)
18
is to reject the null hypotheses (no relationship and conclude that relationship exist
between at least one independent variable with dependent variable)
iii) Forecast with 95% confidence interval the per capita consumption of mutton in 2010.
Assuming that PMT has increased 6% ,PBF has increased 4.5% and INC increased 2.5%
from 2002 to 2010.
Y' ± 2Se
0.825 ±
2(0.0847)
0.6556 < Y' < 0.9944
The adjusted coefficient of determination is 0.719. It means that 71.9% the equation is
explained by the multi linear relationship between per capita consumption of mutton
and its independent variables.
QMT2010= 0.425-0.040(16.64)+0.029(15.47)+0.00004(15425.23)
QMT2010= 0.825
Per Capita Consumption of mutton forecasted to be 0.83kg on year 2010.
6) Conclusion
19
Therefore, differences between Qualitative and Quantitative
Technique
Qualitative Quantitative
Involve mathematical
-
techniques
Example:
Example:
Forecasting newly
sales of colour TVs
introduced online sales
7) References
1. Ferguson, Charles E. (1972). Microeconomic Theory (3rd ed.). Homewood, Illinois: Richard D.
Irwin
3. Wall, Stuart; Griffiths, Alan (2008). Economics for Business and Management. Financial Times
Prentice Hall.
4. Mei-Yu Wang,(2001 ) The Document Supply industry : Supply and demand , MCB University
Press , Vol 50,No 3,2001-Pg 128-131
5. Smith , Todd , Jay Mabe and Jeff Beech,”Components of Demand Planning ,”Strategic
Supply Chain Alignment,Aldershot ,England : Gower Publishing Limited,1998 , pg 123-137
20
6. Smaros, J., Lehtonen, J.M., Appelqvist, P. and Holmstrom, J. (2003), “The impact of increasing
demand visibility on production and inventory control efficiency”, International journal of
Physical Distribution & Logistics Management, Vol. 33 No. 4, pp. 336-543
21