OPERATIONAL LEVEL

Subject Management Accounting (P1)

K. Sivagar
Lecturer B.Sc (Hons), ACMA, CGMA

Module Tutorial 08 - Forecasting Techniques

Code P1/KS/08
 Forecasting Techniques

The purpose of forecasting in the budgeting process is to establish realistic assumptions for planning.
Forecasts might also be prepared on a regular basis for the purpose of feedforward control reporting.

A forecast might be based on simple assumptions, such as a prediction of a 5% growth in sales volume or
sales revenue. Similarly, budgeted expenditure might be forecast using a simple incremental budgeting
approach, and adding a percentage amount for inflation on top of the previous year’s budget.

On the other hand, forecasts might be prepared using a number of forecasting models, methods or
techniques. The reason for using these models and techniques is that they might provide more reliable
forecasts.

There are there forecasting techniques explained in this chapter.

1. The High-Low Method

2. Regression Analysis
3. Time Series Analysis

The High-Low Method

This is a method of breaking semi-variable costs into their two components. A semi-variable cost being a
cost which is partly fixed and partly variable.

Practice Question – 01

X Ltd has had the following output and cost results for the last 4 years.

Output units Cost

$
Year 1 5,000 26,000
Year 2 7,000 34,000
Year 3 9,000 42,000
Year 4 10,000 46,000

In Year 5 the output is expected to be 13,000 units. Calculate the expected costs.

Inflation may be ignored.

P1 2021 – Tutorial 08 – K. Sivagar 1

The high-low method with stepped fixed cost

Sometimes fixed costs are only fixed within certain levels of activity and increase in steps as activity
increases (i.e. they are stepped fixed costs). The high/low method can still be used to estimate fixed and
variable costs. Simply choose two activity levels where the fixed cost remains unchanged.

Adjustments need to be made for the fixed costs based on the activity level under consideration.

Practice Question – 02

An organisation has the following total costs at three activity levels:

Activity level (units) 4,000 6,000 7,500

Total cost $40,800 $50,000 $54,800

Variable cost per unit is constant within this activity range and there is a step up of 10% in the total fixed
costs when the activity level exceeds 5,500 units.

What is the total cost at an activity level of 5,000 units?

P1 2021 – Tutorial 08 – K. Sivagar 2

Regression Analysis

The high-low method only takes account of two observations – the highest and the lowest. To take
account of all observations a more advanced calculation is used known as linear regression which uses a
formula to estimate the linear relationship between the variables as follows:

The equation of a straight line is:

y = a + bx
y = independent variable
a = intercept (on y axis)
b = gradient
x = independent variable

n∑xy− ∑x∑y
b=
n∑x2−(∑x)²

where n = number of pairs of data

a = y̅ - bx̅
x̅ is the arithmetic mean (or average) of x and is calculated as:

∑𝑥
x̅ =
𝑛

y̅ is the arithmetic mean (or average) of y and is calculated as:

∑𝑦
y̅ =
𝑛

P1 2021 – Tutorial 08 – K. Sivagar 3

Practice Question – 03

Marcus Aurelius is a small supermarket chain that has 6 shops. Each shop advertises in their local
newspapers and the marketing director is interested in the relationship between the amount that they
spend on advertising and the sales revenue that they achieve. She has collated the following information
for the 6 shops for the previous year:

(a) Calculate the line of best fit using regression analysis.

Marcus Aurelius has just taken on 2 new stores in the same area and the predicted advertising expenditure
is expected to be $150,000 for one store and $50,000 for the other.

(b) Calculate the predicted sales revenues.

(c) Explain the reliability of the forecasts.

Linear regression analysis can be used to make forecasts or estimates whenever a linear relationship is
assumed between two variables, and historical data is available for analysis.

Linear regression analysis is an alternative to using the high-low method of cost behaviour analysis. It
should be more accurate than the high-low method, because it is based on more items of historical data,
not just a ‘high’ and a ‘low’ value.

– The independent variable (x) in total cost analysis is the volume of activity.

– The dependent variable (y) is total cost.

– The value of a is the amount of fixed costs.

– The value of b is the variable cost per unit of activity.

P1 2021 – Tutorial 08 – K. Sivagar 4

Correlation

Regression analysis attempts to find the linear relationship between two variables. Correlation is
concerned with establishing how strong the relationship is.

Degrees of correlation are:

• Perfect correlation

• Partial correlation

• No correlation

P1 2021 – Tutorial 08 – K. Sivagar 5

Correlation coefficient
• A measure of the amount of linear correlation present in a set of data.
• Pearson’s correlation coefficient, denoted as r and r always lies in the range -1 to +1, where:
✓ r = +1 denotes perfect positive linear correlation
✓ r = -1 denotes perfect negative linear correlation
✓ r = 0 denotes no linear correlation

The coefficient of determination

• The coefficient of determination, denoted as r 2.
• This gives the proportion of changes in y that can be explained by changes in x.

If the correlation coefficient is 0.8, what is the coefficient of determination?

0.64

This means that 64% of the observed changes in Y is due to changes in X but, that 36% of the changes
must be due to other factors.

Limitations of simple linear regression

• Assumes a linear relationship between the variables.

• Only measures the relationship between two variables. In reality the dependent variable is affected
by many independent variables.
• Only interpolated forecasts tend to be reliable. The equation should not be used for extrapolation.
• Regression assumes that the historical behaviour of the data continues into the foreseeable future.
• Interpolated predictions are only reliable if there is a significant correlation between the data.

P1 2021 – Tutorial 08 – K. Sivagar 6

Adjusting forecasts for inflation

The accuracy of forecasting is affected by the need to adjust historical data and future forecasts to allow
for price or cost inflation.

• When historical data is used to calculate a trend line or line of best fit, it should ideally be adjusted to
the same index level for prices or costs. If the actual cost or revenue data is used, without adjustments
for inflation, the resulting line of best fit will include the inflationary differences.

• When a forecast is made from a line of best fit, an adjustment to the forecast should be made for
anticipated inflation in the forecast period.

Practice Question – 04

Production overhead costs at company BW are assumed to vary with the number of machine hours
worked. A line of best fit will be calculated from the following historical data, with costs adjusted to allow
for cost inflation over time.

(a) Reconcile the cost data to a common price level, to remove differences caused by inflation.

(b) If the line of best fit, based on current (20X4) prices, is calculated as:

y = 33,000 + 47x

where y = total production overhead costs in $ and x = the number of machine hours:

calculate the expected total overhead costs in 20X5 if expected production activity is 3,100 machine hours
and the expected cost index is 250.

P1 2021 – Tutorial 08 – K. Sivagar 7

The high low method may be distorted in the presence of inflation. The best technique is to strip out the
inflation, perform the technique as usual and then re-apply the inflation.

Time Series Analysis

A time series is a series of figures recorded over time.

E.g. unemployment over the last 5 years, output over the last 12 months, etc.

The technique of time series is all about looking at the past information and identifying the reasons.
Then using the reasons identified for the forecasting made. Reasons behind past information is known
as components of time series. There are four components.

1. Trend
When looking at the past information there can be a long-term underlying movement and such a
component is known as trend. This is the general direction in which sales are heading. It could be
upwards or downwards.

2. Seasonal variations
These are short-term fluctuations in value due to different circumstances which occur at different times
of the year, different days of the week, different time of day etc.

For example, sales could be more during;

• Every year – New year, Christmas seasons

• Every month – End of the month
• Every week – weekend
• Every day – Lunch time, after work

3. Cyclical variations
These are medium term to long term influences associated with the economy.

For example, economic growth or decline

P1 2021 – Tutorial 08 – K. Sivagar 8

4. Random variations/Residual factors
When looking at the past information, there can also be unforeseen reasons such as natural disasters,
terrorism, sudden change of government. These are normally identified as random variations.

With four reasons identified behind past information any given past information (Y) comprises of four
components.

Based on how components are organized there are two models under time series.

1. Additive model
Y= T +S+ C+ R

2. Multiplicative model
Y= T XS XC X R

But cyclical and random variations are not used in practice due to practical problems. For example,
• The best example for cyclical variation is economic growth. When looking at through out the history
there is no proper relationship identifiable making it impossible to use in forecasting.
• Random variations occur due to unforeseen reasons and if the reason cannot be foreseen forecasting
is not possible.

Therefore, in practice cyclical and random variations are taken as negligible and thereby;

1. Additive model
Y= T +S

2. Multiplicative model
Y= T XS

P1 2021 – Tutorial 08 – K. Sivagar 9

Calculation of the trend
There are three main methods of finding the underlying trend of the data:

(1) Inspection. The trend line can be drawn by eye with the aim of plotting the line so that it lies in the
middle of the data.

(2) Least squares regression analysis. The x axis represents time and the periods of time are numbers,
e.g. January is 1, February is 2, March is 3, etc.

In lease squares regression analysis, Trend is calculated in the form of an equation:

y = a + bx

Where:

y – is the dependent variable (e.g. sales revenue)

x – is the independent variable (e.g. time period)

a & b are constants that will be provided in the question

(3) Moving averages. This method attempts to remove seasonal or cyclical variations by a process of
averaging.

Measuring seasonal variations

Seasonal variation can be calculated using two methods:

1. The Additive model – This is where the seasonal variation is added to/subtracted from the trend to
calculate the forecasted sale.

Y=T+S

Where:

T – is the trend sales value

S – is the seasonal variation

Y – is the total forecasted sale

P1 2021 – Tutorial 08 – K. Sivagar 10

If Y and T are known,

S=Y-T

• The sum of seasonal variations has to be zero in the additive model. If they do not adapt to zero, the
seasonal variations should be adjusted so that they do add up to zero.

2. The Multiplicative model – This is where the seasonal variation is multiplied from the trend to calculate
the forecasted sale.

Y=TxS

Where:
T – is the trend sales value

S – is the seasonal variation (maybe given as a decimal, percentage, percentage change or an index value)

Y – is the total forecasted sale

If Y and T are known,

S = Y/T

P1 2021 – Tutorial 08 – K. Sivagar 11

Practice Question – 05

A business recorded the following quarterly sales for the years 2018 & 2019.

2018: Quarter 1 - $340,000

Quarter 2 - $320,000

Quarter 3 - $390,000

Quarter 4 - $450,000

2019: Quarter 1 - $380,000

Quarter 2 - $345,000
Quarter 3 - $360,000

Quarter 4 - $475,000

Quarter 1 of 2018 is x = 1

The trend equation is: y = 332,143 + 11,190x

(a) Calculate the forecasted trend sale for 2020 Quarter 4 and 2023 Quarter 1

(b) Calculate the seasonal variations of the 4 quarters of 2018, using both the additive & multiplicative
models

P1 2021 – Tutorial 08 – K. Sivagar 12

 Objective Test Questions

Question-01

Question-02

P1 2021 – Tutorial 08 – K. Sivagar 13

Question-03

Question-04

P1 2021 – Tutorial 08 – K. Sivagar 14

Question-05

Question-06

P1 2021 – Tutorial 08 – K. Sivagar 15

Question-07

Question-08

Question-09

P1 2021 – Tutorial 08 – K. Sivagar 16

Question-10

Question-11

P1 2021 – Tutorial 08 – K. Sivagar 17

Question-12

Question-13

P1 2021 – Tutorial 08 – K. Sivagar 18

Question-14

Question-15

P1 2021 – Tutorial 08 – K. Sivagar 19