ACS 800

BUSINESS FORECASTING

ADELEKE, I.A.
Hamadu Dallah
 Course Content
• Forecasting
• Importance of forecasting
• Pattern recognition of risk data
• Probability of future events
• Time plot
• Return plot
 Course Content (Contd.)
• Methods of forecasting
• - Probability analysis
- Regression method
- Time series method
+ Elementary time series
exponential smoothing method
moving average method
semi-average method
method of least squares
 Course Content (Contd.)
• Box-Jenkings Models
- Autoregressive models (AR)
- Moving average models (AR)
- Autoregressive-moving average models
(ARMA)
- Autoregressive-integrated moving average
model (ARIMA)
 Course Content (contd.)

• Volatility forecasting
- ARCH models
- GARCH models
• Simulations
• Reserve forecasting
- Claims reserving
- Credibility rating
• Recommended Texts
(i) Time Series Analysis by Box and Tiao
(ii) Quantitative Techniques in Management
by Vohra, N.D.
(iii) Risk Assessment by Baranoff, Harrington
and Niehaus
(iv) Applied Regression Analysis by
 Learning objectives
After learning the content of this course and
completing the course assignment and case
study, you should be able to:
• Explain why risk forecasts are important to
organisations, risk management professionals,
and the risk management process
• Explain how relevant, complete, consistent,
and organised loss data/risk data are
developed and why such data are important
 Learning objectives
• Describe the risk management significance of
trend analysis and the methods of trending risk
data
• Given a case, analyse past data using appropriate
forecasting technique to project the expected
value of the losses that an organisation will incur
during a given time period
• Define or describe each of the Key terms for this
course
 Understanding forecasting
Definition:
A forecast is an estimate of some future value,
amount, or quantity calculated by
mathematical techniques or determined by
intuition.
• Risk management decisions are based on
information about potential future events
(e.g. loses). The information includes forecasts
of the size (severity) as well as the number of
(frequency) of events.
• The development of accurate forecasts takes
place during the second step of the risk
management process.
Importance of forecasts to organizations, risk management
professionals and risk management process

(1) Organizations use forecasts to make more

informed decisions about future activities
(2) Risk management professionals can use
forecasts to help determine the costs and
benefits of various risk management
techniques
(3) Forecasts are important to the risk
management process because they are vital
to the accuracy of the analysis of the loss
exposures.
 Requirements for forecasts
• To conduct forecasts, risk management
professionals must examine data on past
events. These data are then subjected to
probability and trend analysis to project the
expected value of risks.
•
• To find patterns in past events, risk
management professional must obtain data
that are:
- Relevant
- Complete
- Consistent
- Organised
Example: CALENDAR OF HISTORICAL LOSSES RESULTING FROM
MACHINERY DAMAGE
DATE AMOUNT
21/4/2010 1,004.00
3/5/2010 5,643.00
30/8/2010 7,658.00
4/12/2010 9,876.00
21/4/2011 4,352.00
3/5/2011 7,654.00
30/8/2011 6,785.00
4/12/2011 4,532.00
21/4/2012 2,346.00
3/5/2012 4,325.00
30/8/2012 9,876.00
4/12/2012 5,674.00
21/4/2013 4,532.00
3/5/2013 5,674.00
30/8/2013 4,563.00
4/12/2013 3,452.00
 Amount of Loss Adjusted for Price
Level Changes
• For consistency, historical losses should be
adjusted (or indexed) for price level changes.
Otherwise, the reported amounts of two
physically identical losses occurring in
different years probably will be different.
Inflation makes the later loss appear larger
because it is measured in less valued naira. To
prevent this distortion, risk management
professionals must use loss data expressed in
consistent naira.
• Price indices are used to adjust data so that
they are used in constant naira. A price index
for a year indicates the price of a particular
basket of goods and services relative to some
base year, when the price index for the base
year is set to 100.
Example: Indexing Factor for Adjusting
Losses to Year 4 Naira
Benchmark
Base year year
Year 0 Year 1 Year 2 Year 3 Year 4
Price index 100 115.2 125.9 140.2 148.6

Year Price index Indexing factor

1 P1 P4/P1 1.29
2 P2 P4/P2 1.18
3 P3 P4/P3 1.06
4 P4 P4/P4 1.00
 Adjustment of Historical losses to Year
4 Price Levels
DATE AMOUNT Indexing factor Adjustment Amount Annual Total Annual Frequency
21/4/2010 1,004.00 1.29 1295.16
3/5/2010 5,643.00 1.29 7279.47
30/8/2010 7,658.00 1.29 9878.82
4/12/2010 9,876.00 1.29 12740.04 31,193.49 4
21/4/2011 4,352.00 1.18 5135.36
3/5/2011 7,654.00 1.18 9031.72
30/8/2011 6,785.00 1.18 8006.3
4/12/2011 4,532.00 1.18 5347.76 27,521.14 4
21/4/2012 2,346.00 1.06 2486.76
3/5/2012 4,325.00 1.06 4584.5
30/8/2012 9,876.00 1.06 10468.56
4/12/2012 5,674.00 1.06 6014.44 23,554.26 4
21/4/2013 4,532.00 1.00 4532
3/5/2013 5,674.00 1.00 5674
30/8/2013 4,563.00 1.00 4563
4/12/2013 3,452.00 1.00 3452 18,221.00 4
Total number of losses 16
Total naira losses 100489.89

Mean frequency (per year) = Total number of losses/number of years = 4

Mean severity (per loss) = Total naira losses /total number of losses = N6,280.62
 Probability Analysis
• Probability analysis is a technique for
forecasting events, such as accidental and
business losses, on the assumption that they
are governed by an unchanging probability
distribution.
• Probability analysis is effective for predicting
accidental losses in organisations that have a
substantial volume of data on past losses and
fairly stable operations so that patterns of
past losses presumably will continue
unchanged in the future.
Number Of Rejects From A Production
Line Leading To Loss Of Revenue

Number of rejects Probability

0 0.270
1 0.350
2 0.200
3 0.147
4 0.030
5 0.002
6+ 0.001
 Class Exercise
• The data below represent the number of
stoppages experienced by a Quarry due to
breakdown of crushers in the last 30 months:
• 0, 0, 5, 4, 5, 3, 2, 1, 4, 3, 3, 5, 7, 6, 5, 4, 4, 4, 3,
4, 2, 2, 4, 1, 1, 0, 5, 3, 3, 2.
• The losses experienced as a result of
breakdowns are as given:
Number of
breakdowns 1 2 3 4 5 6 7
Losses (mN) 5 10 23 35 40 45 50
Required
• Present the information in an empirical
probability distribution table
• What are the chances that during the next
production year, there will be
(i) 2 breakdowns
(ii) 4 breakdowns
(iii) up to 3 breakdowns
• For each of (i) to (iii) above, what is the potential
financial loss
Percentage of Percentage of
Number of Total number of Naira Amount of total Naira
breakdowns breakdowns losses amount
•
Assignment
Given the following historical loss data and
price index:
DATE AMOUNT
21/4/2010 1,008.00
3/5/2010 4,651.00
30/8/2010 155.00
4/12/2010 1,783.00
21/4/2011 1,271.00
3/5/2011 6,271.00
30/8/2011 7,119.00
4/12/2011 13,208.00
21/4/2012 5,189.00
3/5/2012 7,834.00
30/8/2012 3,774.00
4/12/2012 12,925.00
19/12/2012 12,830.00
21/4/2013 6,782.00
3/5/2013 21,425.00
3/6/2013 4,483.00
3/7/2013 9,059.00
30/8/2013 4,224.00
4/12/2013 35,508.00
 Year Price index Indexing factor
1 P1 P4/P1 1.29
2 P2 P4/P2 1.18
3 P3 P4/P3 1.06
4 P4 P4/P4 1.00

• Prepare the empirical probability distribution

table using the interval 0-5000, 5001-10000,
10001-15000, 15001-20000, 20001-25000,
25001+
 Solution to Assignment
DATE AMOUNT Indexing factor Adjustment Amount Annual Total Annual Frequency
21/4/2010 1,008.00 1.29 1300.32
3/5/2010 4,651.00 1.29 5999.79
30/8/2010 155 1.29 199.95
4/12/2010 1,783.00 1.29 2300.07 9,800.13 4
21/4/2011 1,271.00 1.18 1499.78
3/5/2011 6,271.00 1.18 7399.78
30/8/2011 7,119.00 1.18 8400.42
4/12/2011 13,208.00 1.18 15585.44 32,885.42 4
21/4/2012 5,189.00 1.06 5500.34
3/5/2012 7,834.00 1.06 8304.04
30/8/2012 3,774.00 1.06 4000.44
4/12/2012 12,925.00 1.06 13700.50
19/12/2012 12,830.00 1.06 13599.80 45,105.12 5
21/4/2013 6,782.00 1.00 6782.00
3/5/2013 21,425.00 1.00 21425.00
3/6/2013 4,483.00 1.00 4483.00
3/7/2013 9,059.00 1.00 9059.00
30/8/2013 4,224.00 1.00 4224.00
4/12/2013 35,508.00 1.00 35508.00 81,481.00 6
Total number of losses 19
Total naira losses 169271.67

Mean frequency (per year) = Total number of losses/number of years = 4.75

Mean severity (per loss) = Total naira losses /total number of losses = N8,909
 Empirical Distributions
• Empirical probability distributions are
estimated from historical data.
• Consider the empirical probability distribution
of data in assignment 1, we have
[1] [2] [3] [4] [5]

Percentage of Percentage of
Size category No. of losses Total losses Amount of Losses Total Amount

0 - 5,000 7 36.84% 18,007.00 10.64%

5,001 - 10,000 7 36.84% 51,445.00 30.39%

10,001 - 15,000 2 10.53% 27,301.00 16.13%

15,001 - 20,000 1 5.26% 15,585.00 9.21%

20,001 - 25,000 1 5.26% 21,425.00 12.66%

25,001 + 1 5.26% 35,508.00 20.98%

Total 19 100.00% 169,271.00 100.00%

 Assignment 2
• The number of defective products from a
production line is an indication of potential losses
of an organization.
• You are required to collect historical data on
number of defective and their monetary values
products from any manufacturing company.
• Classifying the data into the number of bins of
your choice, obtain the empirical probability
distribution of your loss data.
Trend Analysis: Regression Analysis
• Trend analysis is an analysis that identifies
patterns in past risk events and then projects
these patterns into the future
• Regression analysis is a statistical technique that
is used to estimate relationship between
variables.
• Regression analysis is a method of trending risk
data, assuming that the variable being forecast
varies predictably with some other variable(s)
• The variable being forecast is the dependent
variable. The variable that determines the
value of variable being forecast is the
independent variable
• Linear regression analysis is a form of
regression analysis that assumes that the
change in the dependent variable is constant
for each unit change in the independent
variable.
 Example
• Use regression analysis to project trends that
relate the number of machine losses to time
and to annual tons of rubber output as given
below:
TIME SERIES
• Time Series and Time Series Methods
• Components of a Time Series
• Smoothing Methods
• Trend Projection
• Trend and Seasonal Components
• Regression Analysis
• Qualitative Approaches to Forecasting
 Time Series and Time Series
Methods
• By reviewing historical data over time, we can
better understand the pattern of past behavior of a
variable and better predict the future behavior.
• A time series is a set of observations on a variable
measured over successive points in time or over
successive periods of time.
• The objective of time series methods is to discover
a pattern in the historical data and then extrapolate
the pattern into the future.
• The forecast is based solely on past values of the
variable and/or past forecast errors.
Time Series Components

Trend Cyclical

Seasonal Irregular
 The Components of a Time Series
• Trend Component
– It represents a gradual shifting of a time series to
relatively higher or lower values over time.
– Trend is usually the result of changes in the
population, demographics, technology, and/or
consumer preferences.
• Cyclical Component
– It represents any recurring sequence of points
above and below the trend line lasting more than
one year.
– We assume that this component represents
multiyear cyclical movements in the economy.
 The Components of a Time Series

• Seasonal Component
– It represents any repeating pattern, less than one year
in duration, in the time series.
– The pattern duration can be as short as an hour, or
even less.
• Irregular Component
– It is the “catch-all” factor that accounts for the
deviation of the actual time series value from what we
would expect based on the other components.
– It is caused by the short-term, unanticipated, and
nonrecurring factors that affect the time series.
 Trend Component
• Persistent, overall upward or downward
pattern
• Due to population, technology etc.
• Several years duration

Response

Mo., Qtr., Yr. © 1984-1994 T/Maker Co.

 Trend Component
• Overall Upward or Downward Movement
• Data Taken Over a Period of Years
Sales

Time
 Cyclical Component
• Repeating up & down movements
• Due to interactions of factors influencing
economy
• Usually 2-10 years duration

Cycle
Response

Mo., Qtr., Yr.

 Cyclical Component
• Upward or Downward Swings
• May Vary in Length
• Usually Lasts 2 - 10 Years
Sales

Time
 Seasonal Component
• Regular pattern of up & down fluctuations
• Due to weather, customs etc.
• Occurs within one year

Summer
Response

© 1984-1994 T/Maker Co.

Mo., Qtr.
 Seasonal Component
• Upward or Downward Swings
• Regular Patterns
• Observed Within One Year
Sales

Time (Monthly or Quarterly)

 Irregular Component
• Erratic, unsystematic, ‘residual’ fluctuations
• Due to random variation or unforeseen events
– Union strike © 1984-1994 T/Maker Co.

– War
• Short duration &
nonrepeating
 Random or Irregular
Component
• Erratic, Nonsystematic, Random, ‘Residual’
Fluctuations
• Due to Random Variations of
– Nature
– Accidents

• Short Duration and Non-repeating

 Time Series Forecasting
Time
Series

Smoothing No Yes Trend

Methods Trend? Models

Moving Exponential
Average Smoothing

Auto-
Linear Quadratic Exponential Regressive
 Forecast Accuracy
• Mean Squared Error (MSE)
– It is the average of the sum of all the squared
forecast errors.
• Mean Absolute Deviation (MAD)
– It is the average of the absolute values of all the
forecast errors.
One major difference between MSE and MAD
is that the MSE measure is influenced much
more by large forecast errors than by small
errors.
 Outliers

• The purpose of time series analysis is to try to

smooth the data.
• Extreme outliers can distort the estimation of
trend and seasonal components.
• Identifying any outliers and discussing the
effects on other components is important in
your report.

49
 More on outliers
• Definitions:
– An outlier can be defined as an element in the
irregular which is 1.8 standard deviations or
more from the mean.
– An extreme outlier can be defined as one
which is more than 2.8 standard deviations
from the mean.
• Calculate the mean and standard deviation of the
outliers.
• Then identify the values which are outliers by the
definitions on the previous slide.

50
Example
Time Amount
1 1295.16
2 7279.47
3 9878.82
4 12740.04
5 5135.36
6 9031.72
7 8006.3
8 5347.76
mean 6280.618
9 2486.76 Std. dev 3082.303
10 4584.5 outlier limit
= mean + 1.8*std. dev 11828.76
11 10468.56 extreme limit
12 6014.44 = mean + 2.8*std. dev 14911.07

13 4532
14 5674
15 4563
16 3452
 Using Smoothing Methods in
Forecasting
• Moving Averages
– We use the average of the most recent n data
values in the time series as the forecast for the
next period.
– The average changes, or moves, as new
observations become available.
– The moving average calculation is

Moving Average = (most recent n data

values)/n
 Using Smoothing Methods in
Forecasting
• Weighted Moving Averages
– This method involves selecting weights for each of the
data values and then computing a weighted mean as
the forecast.
– For example, a 3-period weighted moving average
would be computed as follows.

Ft + 1 = w1(Yt - 2) + w2(Yt - 1) + w3(Yt)

where the sum of the weights (w values) is 1.

 Using Smoothing Methods in
Forecasting
• Exponential Smoothing
– It is a special case of the weighted moving
averages method in which we select only the
weight for the most recent observation.
– The weight placed on the most recent
observation is the value of the smoothing
constant, a.
– The weights for the other data values are
computed automatically and become smaller
at an exponential rate as the observations
become older.
 Using Smoothing Methods in
Forecasting
• Exponential Smoothing
Ft + 1 = aYt + (1 - a)Ft

where Ft + 1 = forecast value for period t + 1

Yt = actual value for period t + 1
Ft = forecast value for period t
a = smoothing constant (0 < a < 1)
 Example:
A company specializes in conducting management
development seminars. In order to better plan
future revenues and costs, management would like
to develop a forecasting model for their “Time
Management” seminar.
Enrollments for the past ten “TM” seminars are:
(oldest) (newest)
Seminar 1 2 3 4 5 6 7 8 9 10
Enroll. 34 40 35 39 41 36 33 38 43 40
 Example: Executive Seminars, Inc.
• Exponential Smoothing
Let a = .2, F1 = Y1 = 34
F2 = aY1 + (1 - a)F1
= .2(34) + .8(34)
= 34
F3 = aY2 + (1 - a)F2
= .2(40) + .8(34)
= 35.20
F4 = aY3 + (1 - a)F3
= .2(35) + .8(35.20)
= 35.16
. . . and so on
 Example:
Seminar Actual Enrollment Exp. Sm. Forecast
1 34 34.00
2 40 34.00
3 35 35.20
4 39 35.16
5 41 35.93
6 36 36.94
7 33 36.76
8 38 36.00
9 43 36.40
10 40 37.72
11 Forecast for the next seminar = 38.18
 Using Trend Projection in
Forecasting
• Equation for Linear Trend
Tt = b 0 + b 1 t
where
Tt = trend value in period t
b0 = intercept of the trend line
b1 = slope of the trend line
t = time
Note: t is the independent variable.
 Using Trend Projection in
Forecasting
• Computing the Slope (b1) and Intercept (b0)

b1 = tYt - (t Yt)/n

t 2 - (t )2/n

b0 = (Yt/n) - b1t/n = Y - b1t

where
Yt = actual value in period t
n = number of periods in time series
 Example:

Consider a major marine dealer in a country. The

firm has experienced tremendous sales growth in
the past several years. Management would like
to develop a forecasting method that would
enable them to better control inventories.
The annual sales, in number of boats, for one
particular model for the past five years are:

Year 1 2 3 4 5
Sales 11 14 20 26 34
 Example:
• Linear Trend Equation

t Yt tYt t2
1 11 11 1
2 14 28 4
3 20 60 9
4 26 104 16
5 34 170 25
Total 15 105 373 55
 Example:

• Trend Projection
b1 = 373 - (15)(105)/5 = 5.8
55 - (15)2/5
b0 = 105/5 - 5.8(15/5) = 3.6
Tt = 3.6 + 5.8t
T6 = 3.6 + 5.8(6) = 38.4
Quadratic Time-Series Model
Relationships
Y b11 > 0 Y b11 > 0

Year, X 1 Year, X 1
Y b11 < 0 Y b11 < 0

Year, X 1 Year, X 1
 Quadratic Trend Model

Year Coded Sales Ŷ i  b 0  b1 X i  b 2 X i

2
94 0 2
95 1 5 Coefficients
96 2 2 I n te rc e p t 2.85714286
97 3 2 X V a ri a b l e 1 -0 . 3 2 8 5 7 1 4
98 4 7 X V a ri a b l e 2 0.21428571
99 5 6 Excel Output

Ŷ i  2 . 857  0 . 33 X i  . 214 X i
2
Exponential Time-Series Model
Relationships

Y b1 > 1

0 < b1 < 1
Year, X 1
 Exponential Trend Model
Ŷ i  b 0 b 1 or log Ŷ i  log b 0  X 1 log b 1
Xi

Year Coded Sales

C o e f f ic ie n t s
94 0 2
In t e rc e p t 0 .3 3 5 8 3 7 9 5
95 1 5
X V a ri a b l e 10 . 0 8 0 6 8 5 4 4
96 2 2
97 3 2 Excel Output of Values in logs

98 4 7 a n t ilo g (. 3 3 5 8 3 7 9 5 ) = 2.17
99 5 6 a n t ilo g (. 0 8 0 6 8 5 4 4 ) = 1.2

Ŷ i  ( 2 . 17 )( 1 . 2 ) Xi
 Trend and Seasonal Components
in Forecasting
• Multiplicative Model
• Calculating the Seasonal Indexes
• Deseasonalizing the Time Series
• Using the Deseasonalizing Time Series
to Identify Trend
• Seasonal Adjustments
• Cyclical Component
 Multiplicative Model

• Using Tt , St , and It to identify the trend, seasonal, and

irregular components at time t, we describe the time
series value Yt by the following multiplicative time
series model:

Yt = Tt x St x It
• Tt is measured in units of the item being forecast.
• St and It are measured in relative terms, with values
above 1.00 indicating effects above the trend and
values below 1.00 indicating effects below the trend.
 Calculating the Seasonal Indexes
1. Compute a series of n -period centered
moving averages, where n is the number of
seasons in the time series.
2. If n is an even number, compute a series of 2-
period centered moving averages.
3. Divide each time series observation by the
corresponding centered moving average to
identify the seasonal-irregular effect in the
time series.
4. For each of the n seasons, average all the
computed seasonal-irregular values for that
season to eliminate the irregular influence and
obtain an estimate of the seasonal influence,
called the seasonal index, for that season.
 Deseasonalizing the Time Series
• The purpose of finding seasonal indexes is to
remove the seasonal effects from the time
series.
• This process is called deseasonalizing the time
series.
• By dividing each time series observation by
the corresponding seasonal index, the result is
a deseasonalized time series.
• With deseasonalized data, relevant
comparisons can be made between
observations in successive periods.
 Using the Deseasonalizing Time Series
to Identify Trend
• To identify the linear trend, we use the linear
regression procedure covered earlier; in this
case, the data are the deseasonalized time
series values.
• In other words, Yt now refers to the
deseasonalized time series value at time t and
not to the actual value of the time series.
• The resulting line equation is used to make
trend projections, as it was earlier.
 Seasonal Adjustments

• The final step in developing the forecast is to

use the seasonal index to adjust the trend
projection.
• The forecast for period t, season s, is obtained
by multiplying the trend projection for period
t by the seasonal index for season s.

Yt,s = Is[b0 + b1(t )]

Example: Eastern Athletic Supplies
Management of EAS would like to develop
a quarterly sales forecast for one of their
tennis rackets.
Sales of tennis rackets is highly seasonal and
hence an accurate quarterly forecast could
aid substantially in
ordering raw material used in anufacturing.
The quarterly sales data (000 units) for the
previous three years is shown on the next
slide.
Example: Eastern Athletic Supplies
Year Quarter Sales
1 1 3
2 9
3 6
4 2
2 1 4
2 11
3 8
4 3
3 1 5
2 15
3 11
4 3
Example: Eastern Athletic Supplies
Year Quarter Sales 4-CMA 2-CMA
1 1 3
2 9 5.00
3 6 5.25
5.13
4 2 5.75
5.50
2 1 4 6.25
6.00
2 11 6.50
6.38
3 8 6.75
6.63
4 3 7.75
7.25
3 1 5 8.50
8.13
2 15 8.50
8.50
3 11
4 3
Example: Eastern Athletic Supplies
Year Quarter Sales 2-CMA Seas-Irreg
1 1 3
2 9
3 6 5.13 1.17
4 2 5.50 0.36
2 1 4 6.00 0.67
2 11 6.38 1.72
3 8 6.63 1.21
4 3 7.25 0.41
3 1 5 8.13 0.62
2 15 8.50 1.76
3 11
4 3
Example: Eastern Athletic Supplies
Quarter Seas-Irreg Values Seas. Index
1 0.67, 0.62 0.65
2 1.72, 1.76 1.74
3 1.17, 1.21 1.19
4 0.36, 0.41 0.39
Total = 3.97
Seas.Index Adj. Factor Adj.Seas.Index
0.65 4/3.97 .655
1.74 4/3.97 1.753
1.19 4/3.97 1.199
0.39 4/3.97 .393
Total = 4.000
Example: Eastern Athletic Supplies
Year Quarter Sales Seas.Index Deseas.Sales
1 1 3 .655 4.58
2 9 1.753 5.13
3 6 1.199 5.00
4 2 .393 5.09
2 1 4 .655 6.11
2 11 1.753 6.27
3 8 1.199 6.67
4 3 .393 7.63
3 1 5 .655 7.63
2 15 1.753 8.56
3 11 1.199 9.17
4 3 .393 7.63
 Example: Eastern Athletic Supplies
• Trend Projection

Tt = 4.066 + .3933t
T13 = 4.066 + .3993(13) = 9.1789

Using the trend component only, we

would forecast sales of 9,179 tennis rackets
for period 13 (year 4, quarter 1).
 Example: Eastern Athletic Supplies

• Seasonal Adjustments

Period Trend Seasonal Quarterly

t Forec. Index Forecast
13 9,179 .655 6,012
14 9,572 1.753 16,780
15 9,966 1.199 11,949
16 10,359 .393 4,071
 Models Based on Monthly Data
• Many businesses use monthly rather than
quarterly forecasts.
• The preceding procedures can be applied with
minor modifications:
– A 12-month moving average replaces the 4-
quarter moving average.
– 12 monthly, rather than 4 quarterly, seasonal
indexes must be computed.
– Otherwise, the procedures are identical.
 Cyclical Component
• The multiplicative model can be expanded to
include a cyclical component that is expressed as
a percentage of trend.
Yt  Tt  Ct  St  I t
• However, there are difficulties in including a
cyclical component:
– A cycle can span several (many) years and enough
data must be obtained to estimate the cyclical
component.
– Cycles usually vary in length.
 Regression Analysis
• One or more independent variables can be used
to predict the value of a single dependent
variable.
• The time series value that we want to forecast is
the dependent variable.
• The independent variable(s) might include any
combination of the following:
– Previous values of the time series variable itself
– Economic/demographic variables
– Time variables
 Regression Analysis
• An autoregressive model is a regression model
in which the independent variables are
previous values of the time series being
forecast.
• A causal forecasting model uses other time
series related to the one being forecast in an
effort to explain the cause of a time series’
behavior.
 Regression Analysis
• For a function involving k independent variables, we
use the following notation:

Yt = value of the time series in period t

x1t = value of independent variable 1 in period t
x2t = value of independent variable 2 in period t

xkt = value of independent variable k in period t

 Regression Analysis
• In forecasting sales of refrigerators, we might
select the following five independent variables:
x1t = price of refrigerator in period t
x2t = total industry sales in period t - 1
x3t = number of new-house building permits
in period t - 1
x4t = population forecast for period t
x5t = advertising budget for period t
 Regression Analysis
• The n periods of data necessary to develop the
estimated regression equation would appear as:
Period Time Series Value of Independent
Variables
(t) (Yt) (x1t) (x2t) (x3t) . . (xkt)
1 Y1 x11 x21 x31 . . xk1
2 Y2 x12 x22 x32 . . xk2
. . . . . . . .
. . . . . . . .
n Yn x1n x2n x3n . . xkn
 Qualitative Approaches to
Forecasting
• Delphi Method
– It is an attempt to develop forecasts through
“group consensus.”
– The goal is to produce a relatively narrow spread
of opinions within which the majority of the panel
of experts concur.
• Expert Judgment
– Experts individually consider information that they
believe will influence the variable; then they
combine their conclusions into a forecast.
– No two experts are likely to consider the same
information in the same way.
 Qualitative Approaches to
•
Forecasting
Scenario Writing
– This procedure involves developing several
conceptual scenarios, each based on a well-
defined set of assumptions.
– The decision maker must decide how likely each
scenario is and then make decisions accordingly.
• Intuitive Approaches
– A committee or panel seeks to develop new ideas
or solve complex problems through a series of
“brainstorming sessions.”
– Individuals are free to present any idea without
being concerned about criticism or relevancy.
 ARIMA PROCESSES

II. Autoregressive Process

The autoregressive process of order p, AR(p) is
𝑥𝑡 = 𝛼1 𝑥𝑡−1 + 𝛼2 𝑥𝑡−2 + ⋯ + 𝛼𝑝 𝑥𝑡−𝑝 + 𝑢𝑡
If the disturbance is white noise, that is normally
distributed, equation 10 is a pure AR(p) process.
1. Moving Average Process
When the disturbance is not white noise, we assume a moving average
specification, MA(q) process,
𝑢𝑡 = 𝜀𝑡 − 𝛽1 𝜀𝑡−1 − ⋯ − 𝛽𝑞 𝜀𝑡−𝑞
where 𝜀 is a white noise, Eq. 11 specifies a pure MA(q) process.

1. ARMA Process
A mixed of autoregressive and moving average is an ARMA(p, q) process. Thus,
we have
𝑥𝑡 = 𝛼1 𝑥𝑡−1 + ⋯ + 𝛼𝑝 𝑥𝑡−𝑝 + 𝜀𝑡 − 𝛽1 𝜀𝑡−1 − ⋯ − 𝛽𝑞 𝜀𝑡−𝑞
1. ARIMA Process
The non-seasonal ARIMA model as classified as an “ARIMA (p, d, q)” model,
where:
 p is the number of autoregressive terms,
 d is the number of non-seasonal differences, and
 q is the number of lagged forecast errors in the prediction equation.
An Auto Regressive Integrated Moving Average (ARIMA) model predicts
future values of a time series by a linear combination of its past values and a
series of errors 𝜀𝑡 . For a response series 𝑥𝑡 , the general form for the
ARIMA model is:
∅ 𝐵 𝑊𝑡 − 𝑢 = ∅(𝐵)𝜀𝑡

where t is the time index, B is the lag operator defined as𝐵𝑥𝑡 = 𝑥𝑡−1 , 𝑊𝑡 =
1 − 𝐵 𝑑 𝑥𝑡 is the response series after differencing and µ is the intercept or
mean term.
II. PROPERTIES OF AS, MA, AND ARMA PROCESSES
a. AR(1) Process
The specification of the model is
𝑥𝑡 = 𝑚 + 𝛼𝑥𝑡−1 + 𝜀𝑡
The expected value is
𝑚
𝐸 𝑥𝑡 = =𝑢
1−𝛼

this has a constant conditional mean

 The variance can be easily derived as
𝜎𝜀2
𝛾0 = 𝜎𝑥2 =
1−𝛼 2

 The autocovariance coefficients are given by 𝛾1 = 𝛼𝛾0

Similarly
𝛾2 = 𝛼𝛾1 , and in general 𝛾𝑘 = 𝛼𝛾𝑘−1 = 𝛼 𝑘 𝛾0
The autocorrelation coefficients for stationary series are defined by
𝛾𝑘
𝜌𝑘 =
𝛾0

For AR(1) process, the autocorrelation is 𝜌𝑘 = 𝛼𝜌𝑘−1 = 𝛼 𝑘

 a. AR(2) Process
The AR(2) process is defined as
𝑥𝑡 = 𝑚 + 𝛼1 𝑥𝑡−1 + 𝛼2 𝑥𝑡−2 + 𝜀𝑡
The unconditional mean is
𝑚
𝑢=
(1−𝛼1 −𝛼2 )

The autocovariances are

𝛾0 = 𝛼1 𝛾1 + 𝛼2 𝛾2 + 𝜎𝜀2
𝛾1 = 𝛼1 𝛾0 + 𝛼2 𝛾1
𝛾2 = 𝛼1 𝛾1 + 𝛼2 𝛾0
Thus, the variance is
1−𝛼2 𝜎𝜀2
𝛾0 =
1+𝛼2 1−𝛼1 −𝛼2 1+𝛼1 −𝛼2

The autocorrelation coefficients are

𝛼1 𝛼12
𝜌1 = , 𝜌2 = + 𝛼2
1−𝛼2 1−𝛼2

The acf for the AR(2) process is

𝜌𝑘 = 𝛼1 𝜌𝑘−1 + 𝛼2 𝜌𝑘−2 𝑘 = 3, 4, …
 a. MA Processes
The MA(1) process is
𝑥𝑡 = 𝜀𝑡 − 𝛽1 𝜀𝑡−1
The autocovariances are:
𝛾0 = 1 + 𝛽12 𝜎𝜀2 , 𝛾1 = −𝛽1 𝜎𝜀2, 𝛾2 = 𝛾3 = ⋯ = 0
which gives the autocorrelation coefficients as
−𝛽1
𝜌1 = , and 𝜌2 = 𝜌3 = ⋯ = 0
1+𝛽12

a. ARMA Processes
The lowest order mixed process is ARMA (1, 1), it is given as
𝑥𝑡 = 𝛼𝑥𝑡−1 + 𝜀𝑡 − 𝛽𝜀𝑡−1
After taking expectations of Eq. 15 with little manipulation, we find
1−2𝛼𝛽+𝛽 2 2
𝜎𝑥2 = 𝛾0 = 𝜎𝜀
1−𝛼 2
𝛼−𝛽 1−𝛼𝛽
Multiplying Eq.15 by 𝑥𝑡−1 , then taking expectations yields 𝛾1 = 𝛼𝛾0 − 𝛽𝜎𝜀2 = 𝜎𝜀2
1−𝛼 2

Higher autocovariances are given by 𝛾𝑘 = 𝛼𝛾𝑘−1 , 𝑘 = 2,3, …

The autocorrelation function (acf) of the ARMA(1, 1) process is thus
𝛼−𝛽 1−𝛼𝛽
𝜌1 = , 𝜌𝑘 = 𝛼𝜌𝑘−1 𝑘 = 2, 3, …
1−2𝛼𝛽+𝛽 2
III.ESTIMATION METHODS
Software packages (e,g SPSS, STATA)typically offer least-squares, whether linear
or non-linear, or maximum likelihood, whether conditional or full(unconditional),
estimation procedures for ARMA models. Specifically, for AR (1) model given
as
𝑦𝑡 = 𝑚 + 𝛼𝑦𝑡−1 + 𝜀𝑡
where 𝜀 is white noise, the conditional log-likelihood is
𝑛−1 1 𝑛
𝐿𝐿𝑐 = 𝑙𝑛𝐿𝑐 = 𝑐 − 𝑙𝑛𝜎 2 − 𝑡=2 𝑦𝑡 − 𝑚 − 𝛼𝑦𝑡−1 2
2 2𝜎2
The full ML estimates must maximize the unconditional likelihood. It is given by
𝐿𝐿 = 𝑙𝑛𝐿
𝑛 1 1 − 𝛼 2 𝑚 2
= 𝑐 − 𝑙𝑛𝜎 2 + 𝑙𝑛 1 − 𝛼 2 − 𝑦1 −
2 2 2𝜎 2 1−𝛼
𝑛
1
− 2 𝑦𝑡 − 𝑚 − 𝛼𝑦𝑡−1 2
2𝜎
𝑡=2
Taking first derivatives with respect to 𝑚 and 𝛼 no longer yields linear equations
in these parameters; therefore, iterative techniques are required to maximize the
𝐿𝐿 which are often very complex.
 Model Performance Evaluation Criteria
In addition to the𝐿𝐿, we have model estimation evaluation criteria, Akaike Information
Criterion (AIC), Swartz Bayesian Criterion (SBC) and R-Squared. As usual, these criteria
are computed using the log-likelihood estimates under Gaussian assumptions Thus, let
r, k, n and 𝑙𝑙 be response variable, the number parameters, the number of observations
and the maximised likelihood function respectively. Akaike (1978) introduced an
information criterion by penalising the log-likelihood and it is given as:
𝐿𝐿 2𝑘
𝐴𝐼𝐶 = −2 𝑛
+ 𝑛

The Schwarz criterion is an alternative to the AIC that imposes a larger penalty for
additional coefficients. It is given as:
𝐿𝐿 𝑘𝑙𝑛 𝑛
𝑆𝐵𝐶 = −2 +
𝑛 𝑛
R-Squared
This is the usual coefficient of determination
 Case Studies
• Data from CBN and other sources would be
used to illustrate the application of the
concepts explained in this materials.
Claims Forecasting
Assignment
The Table below shows the malicious damage claim for a collection of policies in
successive development years.
i. Adjusted below using inflation of 4%, 2% and 4% for 2002, 2003 and 2004
respectively.
ii. Estimate the cumulative incurred claims in 2004 using chain ladder method with
inflation

Development year
Year of origin 0 1 2 3
2001 2,144 366 234 165
2002 2,231 340 190
2003 2,335 270
2004 2,392
End