Linear Regression

In order to prepare budgets, forecasts of costs and revenues will need to be undertaken.
Regression analysis attempts to find the factors that bring about the change in results. For
example, what is the relationship between advertising expenditure and sales volume?
Linear regression can be used to determine the relationship between two variables
and then to forecast other levels of costs / sales

Variables
The dependent variable ("y") is influenced by the independent variable ("x"). A linear
relationship
y = a + bx
will be established with "a" being the value of "y" when "x" is equal to 0. (eg what would
sales be with no advertising expenditure?) and "b" the increase/(decrease) in "y" brought
about by a change in x by 1 unit (eg how much extra sales revenue would Rs 1 of advertising
expenditure generate?).

Example
Quarter 1 2 3 4 5 6 7 8
Advertising expt (Rs ’000) 22 15 7 44 29 18 10 37
Sales (‘000 units) 11 12 7 20 14 11 11 22

To calculate the values of a and b for the straight line expression y = a + bx, the following
formulae are used:

Required
(a) Calculate b and a, and the expression for sales.
(b) Forecast the level of sales if advertising is Rs 20,000.
Ans y = 5.415 + 0.3554x 12,523 units

Q1 Following information is extracted from the records of Mega Limited for the year ended
Aug June 30, 2014:
2014 Direct Labour Electricity Cost
Month (Hours ‘000’) (Rs. ‘000’)
July 2013 34 640
August 30 620
September 34 620
October 39 590
November 42 500
December 32 530
Jan 2014 26 500
February 26 500
March 31 530
April 35 550
May 43 580
June 48 680
1
 Required:
Compute variable electricity rate and fixed cost per month under the least square
(simple regression) method : 08

Q2 The management of Good Luck Company has asked for help in selection of the appropriate
Sept activity measures to be used in estimating electricity cost while preparing budget for one of its
2013 plants at Lahore. The information as given below shows utility expenses incurred in the past
year with two potential activity measures.

Month Utility Cost Machine Labour

(Rs.) Hours Hours
January 160,000 2,300 4,200
February 157,000 2,250 4,000
March 161,000 2,400 4,360
April 155,000 2,250 4,000
May 153,000 2,160 4,050
June 154,000 2,240 4,100
July 152,000 2,180 4,150
August 153,000 2,170 4,250
September 158,000 2,260 4,150
October 165,000 2,500 4,500
November 166,000 2,540 4,600
December 162,000 2,450 4,400
1,896,000 27,700 50,760

Required:
(a) Compute the coefficient of correlation ‘r’ and the coefficient of determination ‘r2’ between
the cost of utility and each of the two activity measures. 08
(b) Identify which of the two activity measures should be used as a basis to estimate the
allowable cost of utility. 03
(c) Using the activity measure selected in requirement (b) above, compute an estimate of
fixed utility cost and the variable utility rate by the method of least squares. 04

2
 Time series analysis
An analysis of past patterns of demand or sales which will be used to construct expected
patterns in the future.
Components of a time series
(a) Trend (T): the underlying increase or decrease in demand.
For example, a steady decline in the average sales of a national daily newspaper or a
steady increase in sales of Sony PlayStations.
(b) Seasonal variations (SV): short-term repeated fluctuations from the trend.
For example, sales of tabloid newspapers being higher on Mondays and Saturdays
than other days due to the extra sports coverage, or sales of ice cream being higher
in summer than in winter.
(c) Cyclical variations: recurring patterns over a longer period of time, not generally of a
fixed nature.
For example, changes in unemployment, movement from recession to economic growth.
(d) Random variations: these will be included in past data but could not be included in
future estimates.
For example, high sales of a tabloid newspaper due to exclusive photographs of a
member of the royals in a compromising position.

The additive model and the multiplicative (proportional)model

In the additive model the components are assumed to add together to give the time series
TS = T + SV
The multiplicative model multiplies the components together.
TS = T x SV
The multiplicative model is better when the trend is increasing or decreasing over time.

Trend and seasonal variation

The trend can be found using moving averages.
If seasonal variations repeat after four periods, say, we can smooth out these variations,
and calculate the underlying trend, by looking at averages for each quarter’s figures.
Then each average can be compared to an actual value and the seasonal variation found:
Additive model: SV = TS – T
Multiplicative model: SV = TS
T
Seasonally adjusted/deseasonalised data
Seasonal variation factors can be used to smooth actual data.
Additive model : Deseasonalised sales = Actual sales – Seasonal variation
Multiplicative model : Deseasonalised sales = Actual sales
Seasonal variation

Ex-1 The following represents Trisaris Ltd's sales figures and trend figures over the last three years.
Time Period Time Series (sales)
Y1 Q1 18
Q2 60
Q3 90
Q4 102

3
 Y2 Q1 30
Q2 72
Q3 99
Q4 120
Y3 Q1 36
Q2 90
Q3 114
Q4 135
Required
Use the multiplicative model to forecast quarterly sales in Year 4.

Ex-2 The management accountant at Ran Bay Sunglasses has determined the trend equation of sales
against time is
y = 1,000,000 + 80,000 x
where x denotes the quarter (x increases by one for each new quarter) and y is unit sales. The
average seasonal variations in demand are shown below:
Q1 Q2 Q3 Q4
Average S.V. -15% 10% 15% -10%
Required
Forecast sales for quarters 17 to 20, assuming the same pattern of seasonal variations to apply.

Q1 Al-Asar International manufactures and sales non-carbolic drinks. Demand for product is
Feb increasing approximately 10% per annum, but vary based on climatic conditions in
2014 different seasons of year. Quarterly sales data for last two years is as under:
Year Quarter Volume of Sales
(Million Bottles)
2012 Q-1 450
Q-2 750
Q-3 825
Q-4 625
2013 Q-1 1500
Q-2 825
Q-3 900
Q-4 675
Required:
You have been working as Financial Controller in Al-Asar International and asked by
managing partner to calculate the following:
(i) Quarter-wise trend (T) for 2012-13 (Q-3, Q-4, Q-1 and Q-2). 07
(ii) Seasonal variances (SV) for above quarters using Proportional (Multiplicative) Model
(Answer T and SV of Q2 2013 - 718.75 & 1.148)

Q2 The Western is a local government organisation responsible for waste collection

from domestic households. The new management accountant of The Western has
decided to introduce some new forecasting techniques to improve the accuracy of the
budgeting. The next budget to be produced is for the year ended 31 December 2010.
4
 Waste is collected by the tonne (T). The number of tonnes collected each year has been
rising and by using time series analysis the new management accountant has produced
the following relationship between the tonnes collected (T) and the time period in question
Q (where Q is a quarter number. So Q = 1 represents quarter 1 in 2009 and
Q = 2 represents quarter 2 in 2009 and so on)
T = 2,000 + 25Q
Each quarter is subject to some seasonal variation with more waste being collected in the
middle quarters of each year. The adjustments required to the underlying trend prediction are:
Quarter Tonnes
1 -200
2 +250
3 +150
4 -100
Once T is predicted the new management accountant hopes to use the values to predict
the variable operating costs and fixed operating costs that The Western will be subjected
to in 2010. To this end he has provided the following operating cost data for 2009.
Total operating cost in 2009
Volume of waste (fixed + variable)
Tonnes $’000s
2,100 950
2,500 1,010
2,400 1,010
2,300 990
Inflation on the operating cost is expected to be 5% in 2010.
The regression formula is shown on the formula sheet.
Required:
(a) Calculate the tonnes of waste to be expected in the calendar year 2010. (6 marks)
(b) Calculate the variable operating cost and fixed operating cost to be expected in 2010
using regression analysis on the 2009 data and allowing for inflation as appropriate.
(6 marks)
c) Using your answer of (a) and (b), calculate quarterly operating cost of waste
collection for 2010. (6 marks)
(Answer a) 8,750 tonnes b) V168 per ton F 648,900 c))

Q3 The MD of a large furniture store, engaged a management accountant to devise a simple

and practical method of forecasting the store's quarterly sales level for a period of six
months ahead. For this task he gathered the store's gross quarterly sales of last 30 months.
Sales period Value of retail sales
2014 Q1 2,850 Rs '000'
Q2 3,100
Q3 3,150
Q4 3,850
2015 Q1 3,400
Q2 3,700
Q3 3,750
Q4 4,600
2016 Q1 3,950
Q2 4,250

5
 The management accountant found following national quarterly seasonal index in a
natiomnal journal. He thought that this store has a product mix not too different from
aggregate mix on which the index was based.
Q1 94
Q2 98
Q3 96
Q4 112
Required:
a) Calculate the values of the deseasonalised data of past ten quarters.
b) Plot the actual sales figures and deseasonlised data on graph paper.
c) Use the method of least square to determine the equation of the straight line through
the deseasonlised date.
d) Estimate the gross sales figures for the third and fourth quarter of 2016.

Q4 Sauce Co manufactures and sells cartons of cooking sauces, which deteriorate over time and must
be used within three months. Over the last two years, Sauce Co has experienced all kinds of problems.
The financial and sales directors believe these to be a result of persistently unrealistic sales targets
imposed by the managing director, who makes forecasts based on his own subjective and overly
optimistic views about future sales. Whilst an incentive scheme is in place for employees, the company
has not hit its targets for the last three years, so no bonuses have been paid out. The financial director
has asked you to forecast the sales for the last two quarters of 2012, hoping to present these figures
to the managing director in an attempt to persuade him that the basis of forecasting needs to be
changed. Production volumes are also currently based on anticipated sales rather than actual orders.

The following sales figures are available for the last two years. All of the figures represent actual sales
except for quarter 2 of 2012, which is an estimate. The financial director is satisfied that this estimate
can be relied upon.
’000 units
Year Quarter One Quarter Two Quarter ThreeQuarter Four
2010 900 1100
2011 1200 1000 1050 1300
2012 1400 1150

The following centred moving averages have been calculated, using a base period of four quarters:
2011 1068.75 1112.5 1162.5 1206.25
2012 1243.75 1287.5

The average seasonal variations for 2010 have already been made available to you and are 0·908
for quarter 3 and 1·082 for quarter 4. The random component is negligible and can therefore be ignored.

Required:
Using the information provided above, and assuming a proportional (multiplicative) model, forecast the sales
of Sauce Co for the last two quarters of 2012. Calculate your seasonal adjustments to four decimal places.
(Answer sales Q1 & Q2 1,331 & 1,375) (10 marks)