ACCA – APPLIED KNOWLEDGE

MANAGEMENT ACCOUNTING
MA Chapter

BISC TRAINING CENTER

Ms. Nguyen Thi Phuong Thao, ACCA
www.bisc.edu.vn
085 8822 168
[email protected]

Chapter 4
FORECASTING

1
 0. CHAPTER 4 – MAIN PARTS
Forecasting

Correlation and The correlation coefficient

Linear regression
statistical forecasting and the coefficient
techniques of determination
Reliability of regression line
Correlation Correlation coefficient r
Advantage & disadvantages
Scatter graphs of linear regression analysis Coefficient of
determination 𝒓𝟐
Sales forecasting:
Time series analysis Index number
the product life cycle

Components of a time series Price indices and

quantity indices
Time series models

Calculating seasonal variations

Advantage & disadvantages of 3

time series analysis

1. CORRELATION AND STATISTICAL

FORECASTING TECHNIQUES
Correlation: Two variables are said to be correlated if a change in the value
of one variable is accompanied by a change in the value of another variable.

For example: Sales and unit sold, the distance and the time it takes

2
 1. CORRELATION AND STATISTICAL
FORECASTING TECHNIQUES
Degree of correlation Perfectly correlated

All the pairs value on a straight line  exact linear relationship

1. CORRELATION AND STATISTICAL

FORECASTING TECHNIQUES
Degree of correlation Partly correlated

• Although there is no exact relationship, low value of X to be

associated with low value of Y or high value of X with high value of Y
• Again, no exact relationship, but low value of X to be associated with
high value of Y and vice versa.

3
 1. CORRELATION AND STATISTICAL
FORECASTING TECHNIQUES
Degree of correlation Uncorrelated

Two variable but not correlated with other

Correlation can be positive (higher value of one variable associates with
higher value of other variable) or negative (higher value of one variable
associates with lower value of other variable)

1. CORRELATION AND STATISTICAL

FORECASTING TECHNIQUES
Scatter graphs
For example: Total costs at different levels of output could be plotted as:

The ‘independent variable’ is plotted on the x axis and is output in this case
Total cost is the ‘dependent variable’ and is plotted on the y axis

4
 2. LINEAR REGRESSION
Linear regression finds an equation for the line of best fit mathematically.
Once an equation for a line of best fit has been determined, forecasts can
be made. The equation represents the trend of the data.

y = a + bX
∑ ∑ ∑
b= ∑ (∑ )
∑ ∑
a = 𝑦 − 𝑏𝑥̅ = −𝑏

∑ means the ‘sum of’

𝑦 means the average of y (ie ∑ 𝑦 ÷ 𝑛 as shown in the formula)
𝑥 means the average of x (ie ∑ 𝑥 ÷ 𝑛 as shown in the formula)

2. LINEAR REGRESSION
Activity: Regression line
Wigwam Co makes high quality tents for outdoor festivals. The company
accountant has observed costs at different production levels as follows:

Output (units) Total costs ($)

280 46,500
350 49,100
200 36,700
160 32,000
240 44,500

10

10

5
 2. LINEAR REGRESSION
These costs could be plotted on a scatter graph as follows:

The costs appear to follow an approximately linear pattern.

Required:
Calculate the regression line and use the line to estimate costs for output of 240
units and 700 units

11

11

2. LINEAR REGRESSION
Answer: Regression line
x (units) y ($’000) xy ($’000) 𝒙𝟐 𝒚𝟐
280 46.5 13,020 78,400 2,162.25
350 49.1 17,185 122,500 2,410.81
200 36.7 7,340 40,000 1,346.89
160 32.0 5,120 25,600 1,024.00
240 44.5 10,680 57,600 1,980.25
1,230 208.8 53,345 324,100 8,924.20

, ( , . ) ,
b= = = 0.092
, , ,
. ,
a= − 0.092 𝑥 = 19.128
y = $19.128 + $92x
Interpolation, when output is 240 units:
Cost predicted = $19,128 + $92 x 240 = $41,208
Extrapolation, when output is 700 units:
Cost predicted = $19,128 + $92 x 700 = $83,528 12

12

6
 2. LINEAR REGRESSION
Reliability of regression line

Interpolation: Interpolation means using a line of best fit to predict a value

within two extreme points of the observed range.
Extrapolation: Extrapolation means using a line of best fit to predict a
value outside the two extreme points of the observed range.

13

13

2. LINEAR REGRESSION

ADVANTAGES  Assumes linearity between x and y

 The observations used may be atypical
 Gives a definitive line of best fit
 Historic data is used and patterns may
 Makes efficient use of data and change in future
good results can be obtained with
relatively small amounts of data  Each observation should be
independent from the others
 Many processes are linear and so
are well described by regression  Forecasting usually involves
analysis extrapolation outside the given range
of observations where working
conditions and therefore cost patterns
may change

DISADVANTAGES
14

14

7
 2. LINEAR REGRESSION
Exercise: A company has recorded its total cost for different levels of
activity over the last five months as follows:
Month Activity level (units) Total cost ($)
7 300 17,500
8 360 19,500
9 400 20,500
10 320 18,500
11 280 17,000

The equation for total cost is being calculated using regression analysis
on the above data. The equation for total cost is of the general from ‘y =
a + bx’ and the value of ‘b’ has been calculated correctly as 29.53

What is the value of ‘a’ in the total cost equation?

15

15

2. LINEAR REGRESSION
Answer:
• The value of ‘a’ in the total cost equation:
∑ 𝑦 = 17,500 + 19,500 + 20,500 + 18,500 + 17,000 = 93,000
∑ 𝑥 = 300 + 360 + 400 + 320 + 280 = 1,660
a = (93,000 ÷ 5) – (29.53 x 1,660 ÷ 5) = 8,796.04

16

16

8
 3. THE CORRELATION COEFFICIENT AND
THE COEFFICIENT OF DETERMINATION
Correlation coefficient r

Correlation coefficient (r): The correlation coefficient (r) measures the

degree of linear correlation between two variables.

Correlation coefficient (r) indicates the strength of the linear relationship

between x and y.

∑ (∑ ∑ )
r=
∑ (∑ ) ∑ (∑ )

Where X and Y represent pairs of data

for two variables X and Y
n = the number of pairs of data used in
the analysis

17

17

3. THE CORRELATION COEFFICIENT AND

THE COEFFICIENT OF DETERMINATION
Correlation coefficient r

-1≤ r ≤ 1
• r = +1 means perfectly positive correlation
• r = -1 means perfectly negative correlation
• r = 0 means uncorrelated

Activity: Correlation coefficient

What is the correlation coefficient for the data in the previous activity
Wigwam Co?

18

18

9
 3. THE CORRELATION COEFFICIENT AND
THE COEFFICIENT OF DETERMINATION
Correlation coefficient r

Answer
x (units) y ($’000) xy ($’000) 𝒙𝟐 𝒚𝟐
280 46.5 13,020 78,400 2,162.25
350 49.1 17,185 122,500 2,410.81
200 36.7 7,340 40,000 1,346.89
160 32.0 5,120 25,600 1,024.00
240 44.5 10,680 57,600 1,980.25
1,230 208.8 53,345 324,100 8,924.20

,
r=
, ( , . . )
,
r= = 0.94
, , .

19

19

3. THE CORRELATION COEFFICIENT AND

THE COEFFICIENT OF DETERMINATION
Coefficient of determination 𝒓𝟐

Coefficient of determination (𝒓𝟐 ): The coefficient of determination 𝒓𝟐

measures the proportion of the total variation in the value of one variable
that can be explained by variations in the other variable. It denotes the
strength of the linear association between two variables.

Coefficient of determination (𝒓𝟐 ) does not prove a cause and effect

relationship. It merely suggests a possible link
0 ≤ 𝒓𝟐 ≤ 1

Example: What is the coefficient of determination for the data in the

activity Wigwam Co?

20

20

10
 4. TIME SERIES ANALYSIS
Time series: A time series is a series of figures or values recorded over time.

The analysis of time series allows historical data to be monitored so that

observations can be made about how a variable has performed over a
period of time.
Examples:
• Output at a factory each day for the last month
• Total costs per annum for last ten years
• Monthly sales over the last five years.

21

21

4. TIME SERIES ANALYSIS

The components of time series
The time series analysis forecasting technique is usually used to forecast
sales
• A graph of a time series as called a histogram: Horizontal axis is always
represent time, vertical axis represents the values of data
Components
• Trend
• Seasonal variations or fluctuation
• Cycles or cyclical variations
• Random variations

22

22

11
 4. TIME SERIES ANALYSIS
Components of a time series (TS)
 Trend (T): underlying long-term movement over time in the values of
the data recorded (eg growth, inflation).
 Seasonal variations (SV): short term fluctuations (year, week, day, or
whatever)
 Cyclical variations (CV): recurring patterns over a longer period of time,
not generally of a fixed nature (ie recession/depression/economic
growth)
 Random variations (RV): irregular/unpredictable variations, due to
rare/chance occurrences (hurricanes, floods, nuclear war)

23

23

4. TIME SERIES ANALYSIS

Time series models

Additive
model
TS = T + SV + CV + RV
TS = T + SV

Multiplicative
model
TS = T x SV x CV x RV
TS = T x SV

24

24

12
 4. TIME SERIES ANALYSIS
Moving averages
The use of moving averages is the main method for calculating a trend
from a time series. The technique averages all of the results of a fixed
number of periods and relates to the mid-point of the overall period.

25

25

4. TIME SERIES ANALYSIS

Moving averages
• A moving average as an average of the results of a fixed number of
periods  related to the mid-point of the overall period
• Moving averages of an odd number: the moving averages were taken
of the results in an odd number of time periods, and the average then
related to the mid-point of the overall period.
• Moving averages of an even number: we need the mid-point related
to a single period  take a moving average of the moving average

26

26

13
 4. TIME SERIES ANALYSIS
Calculating seasonal variations
Once the trend has been established, the seasonal variation can be
determined using the models above.
 Additive model: SV = TS – T; the sum of seasonal variations must equal
zero. Adjustments need to be made where this is not the case. For
examples if the seasonal variations sum to 2, then we spread 2 across
the quarters so that the final total of the variations sum to zero.
• Multiplicative model: SV = TS / T; the sum of the seasonal variations must
equal the number of periods over which seasonality occurs prior to
repeat ie the sum of a quarterly seasonal variation should equal 4. if, for
example, the sum of a quarterly seasonal variation was 4.7, then we
would deduct 0.7 by spreading it across the quarters, ie by deducting
0.7/4 = 0.175 from each one.
27

27

4. TIME SERIES ANALYSIS

Calculating seasonal variations - Additive model

Time series 3 period moving average Seasonal variation

(TS) (T) (SV= TS – T)
2
5 4 1
5 5 0
5 6 -1
8 7 1
8 8 0
8 9 -1
11

28

28

14
 4. TIME SERIES ANALYSIS
Calculating seasonal variations - Multiplicative model
TS= T x SV
SV = TS/T
Actual Trend Seasonal percentage
Year Quarter (TS) (T) (TS/T)
20X5 1 600
2 840
3 420 650.00 0.646
4 720 657.50 1.095
20X6 1 640 660.00 0.970
2 860 662.50 1.298
3 420
4 740

Suppose that seasonal variations for the next four quarters are 0.628, 1.092 0.980
and 1.309 respectively.
The summary of the seasonal variations expressed in proportional terms is
therefore as follows

29

29

4. TIME SERIES ANALYSIS

Calculating seasonal variations - Multiplicative model
Year Q1 Q2 Q3 Q4
20X5 0.646 1.095
20X6 0.970 1.298 0.628 1.092
20X7 0.980 1.309
Total 1.950 2.607 1.274 2.187
Average 0.975 1.3035 0.637 1.0935

Instead of summing to zero, as with the additive approach, the averages should
sum (in this case) to 4.0 which mean 1.0 for each of the four quarters. They
actually sum to 4.009 so 0.00225 has to be deducted from each one.

30

30

15
 4. TIME SERIES ANALYSIS
Calculating seasonal variations - Multiplicative model
Q1 Q2 Q3 Q4
Average 0.9750 1.30350 0.63700 1.09350
Adjustment -0.00225 -0.00225 -0.00225 -0.00225
Final estimates 0.97275 1.30125 0.63475 1.09125
Rounded 0.97 1.3 0.64 1.09

Note that the proportional model is better than the additive model when the
trend is increasing or decreasing over time. In such circumstances, seasonal
variations are likely to be increasing or decreasing too. The additive model simply
adds absolute and unchanging seasonal variations to the trend figures whereas
the proportional model, by multiplying increasing or decreasing trend values by a
constant seasonal variation factor, takes account of changing seasonal variations.

31

31

4. TIME SERIES ANALYSIS

Calculating seasonal variations - Sales forecasting
You are given the regression equation of how the sales trend varies with time:
y= 400 + 20x
where:
x = quarter (x increases by one for each new quarter)
y = unit sales
Required
Forecast sales for year 6 for each quarter using the seasonal variations below.

Year Trend sales SV Forecast sale

Q1 -5%
Q2 +10%
Q3 +20%
Q4 -25%

32

32

16
 4. TIME SERIES ANALYSIS
Calculating seasonal variations - Sales forecasting
You are given the regression equation of how the sales trend varies with time:
y= 400 + 20x
where:
x = quarter (x increases by one for each new quarter)
y = unit sales
Required
Forecast sales for year 6 for each quarter using the seasonal variations below.

Year Trend sales SV Forecast sale

Q1 820 -5% 779
Q2 840 +10% 924
Q3 860 +20% 1032
Q4 880 -25% 660

33

33

4. TIME SERIES ANALYSIS

• Trend lines can be reviewed and assessed after each
period for reliability, possibly leading to improved
Advantages forecasts with experience.
• Time series components and theory is relatively easy
for non-financial managers to understand.
• The further into the future the forecast the more
unreliable it is likely to be
• The less data available on which to base the forecast
the less reliable the forecast.
Disadvantages
• The pattern of trend and seasonal variation cannot
be guaranteed to continue
• There is always the danger of random variations
upsetting the pattern of trend and seasonal variation.

34

34

17
 5. INDEX NUMBERS
Price indices and quantity indices
• A price index measures the change in the money value
Price index = Pn / Po x 100
Where
• Pn is the price for the period under consideration
• Po is the price for the base period
• A quantity index measures the change in the non-monetary value
Quantity index = Qn/Qo x 100
Where
• Qn is the quantity for the period under consideration
• Qo is the quantity for the base period

35

35

5. INDEX NUMBERS
Price indices and quantity indices
Two ways which index relatives can be calculated
• Fixed base method: a base year is selected, and all subsequent
changes are measures against this base  appropriate if the basic
nature of the commodity is unchanged overtime
• Chain base method: base year is changed  appropriate if the basic
nature of the commodity is changing over time.

36

36

18
 5. INDEX NUMBERS
Activity: Cost of living index
Suppose a cost of living index is to be calculated from the from the
following three items: fizzy drink, pizza and chocolate and that prices for
20X1 and 20X2 were as follows:
X1 X2
Fizzy drink $1.50 $1.80
Pizza $3.50 $3.55
Chocolate $0.55 $0.60
What is the cost of living index for X2, assuming X1 as a base year?

37

37

5. INDEX NUMBERS
Answer:

X1 X2
Fizzy drink $2.50 $2.10
Pizza $3.50 $3.55
Chocolate $0.55 $0.60
∑ 𝑃 6.55 ∑ 𝑃 6.25

𝟔.𝟐𝟓
Price index = x 100 = 95.4%
𝟔.𝟓𝟓

Prices have therefore decreased by 4.6% in a year.

38

38

19
 5. INDEX NUMBERS
Base weighted price indices (Laspeyres index)
Laspeyres indices
Use weights from the base period and are therefore sometimes called
base weighted indices.
Laspeyres price index
A Laspeyres price index uses quantities consumed in the base period as
weights.
∑ Pn Q 0
x 100
∑ P0 Q 0
Laspeyres quantity index
A Laspeyres quantity index uses prices from the base period as weights
∑ P0 Q n
Laspeyre quantity index = x 100
∑ P0 Q 0

39

39

5. INDEX NUMBERS
Current weighted indices (Paasche index)
Paasche indices
Use current time period weights. In other words the weights are
changed every time period.
Paasche price index
A Paasche price index uses quantities consumed in the current period as
weights
∑ Pn Qn
Paasche price index = x 100
∑ P0 Q n
Paasche quantity index
A Paasche quantity index uses prices from the current period as weights
∑ Pn Qn
Paasche quantity index = x 100
∑ Pn Q

40

40

20
 5. INDEX NUMBERS
Calculating quantity indices

We have seen that price indices are usually weighted by quantities. We

will now look at quantity indices, which measure changes in quantities
and use prices as weights. One application of this is to measure changes
in national income.
∑
Laspeyres quantity index = ∑ 𝑥 100
where 𝑃 represents the prices in the base year, 𝑄 represents quantities
in the base year and 𝑄 represents quantities using current-year values.

41

41

5. INDEX NUMBERS
Activity: Quantity indices
The following data relates to production in Country D in 20X3 and 20X4

Quantity Price per unit Quantity Price per unit

produced produced
20X3 20X3 20X4 20X4
‘000 $ ‘000 $
Good A 3 1.20 4 1.5
Good B 6 0.95 5 0.98
Good C 1 1.4 2 1.30
Good D 4 1.1 3 1.14

Calculate the following quantity indices for 20X4 (with 2023 as the
base year)
• A quantity index using base year weightings
• A quantity index using current year weightings

42

42

21
 5. INDEX NUMBERS
Answer
Qo Po Qn Pn PoQo PoQn PnQo PnQn
Good A 3 1.20 4 1.5 3.60 4.80 4.50 6.00
Good B 6 0.95 5 0.98 5.70 4.75 5.88 4.90
Good C 1 1.4 2 1.30 1.40 2.80 1.30 2.60
Good D 4 1.1 3 1.14 4.40 3.30 4.56 3.42
15.10 15.65 16.24 16.92

Quantity index numbers for 20X4 are as follows:

• Laspeyres quantity index: (15.65/15.10) x 100 = 103.64
• Paasche quantity index: (16.92/16.24) x 100 = 104.19

43

43

5. INDEX NUMBERS
Which to use – Paasche or Laspeyres?

Paasche Laspeyres
Paasche index may be more costly Because Laspeyres denominator is
as the index requires quantities to fixed, index numbers for several
be ascertained each year different years can be compared
Paasche index denominator must The weights for Laspeyres index
be recalculated each year become out of date
The Paasche price index tends to A Laspeyres price index tends to
understate inflation. overstate inflation.

44

44

22
 5. INDEX NUMBERS
Fisher’s ideal index
Fisher’s ideal index = (𝐿𝑎𝑠𝑝𝑒𝑦𝑟𝑒𝑠 𝑥 𝑃𝑎𝑎𝑠𝑐ℎ𝑒)

Example:
The Laspeyres index of retail prices for 20X7 (with a base year of 20X1) is
150.2. The corresponding Paasche index is 134.9.
Calculate Fisher’s ideal index

45

45

6. SALES FORECASTING: THE

PRODUCT LIFE CYCLE
The product life cycle (PLC) model shows how sales of a product can be
expected to vary with the passage of time. It can be divided into five
stages.

46

46

23
 6. SALES FORECASTING: THE
PRODUCT LIFE CYCLE
Characteristics of the PLC

Stage Sales volume Costs

Development None Research & development
Very high fixed costs (eg non-current assets,
Introduction Very low levels
advertising)
Increase in variable costs
Growth Rapid increase Some fixed costs increase
(eg increase number of non-current assets)
Stable
Maturity Primarily variable costs
High volume
Primarily variable costs (now decreasing)
Decline Falling demand
Some fixed costs (eg decommissioning costs)

47

47

Q&A

48

48

24