DEMAND FORECASTING

1
DEMAND FORECASTING
2
Demand Forecasting

• To get an overview of the market and act
  proactively
• To adjust production and avoid over production
  and under production
• Essential for production scheduling, purchase of
  raw materials, arranging finance and advertising

3

• Process of forecasting DD and sales of a firms
  product
• Firm uses macro forecast of general economic
  activity (GNP) as inputs for micro forecasts

4
Qualitative Methods

• Used for short term forecasts when data are not
  available
• Also for supplementing quantitative forecasts
• Surveys on economic intentions can reveal and can
  be used to forecast future purchase of capital
  equipment, inventory changes and major consumer
  expenditures

5
Survey Methods

• Complete enumeration (census) vs sample
• Questionnaire, interview or observation
• Opinion Polls
• Consumer Survey
• Executive Polling
• Sales force polling

6
Survey Methods

• Buying Intentions of consumers has limited use
  because
• - Consumer may not be able to clearly foresee the
  choice,
• - Wishful thinking
• -Answers tailored to impress interviewer
• - New alternatives may emerge,
• -Passive because it does not measure variables
  which are under management control
• -Intention may not translate into actual buying

7
End-use method

• Steps
• Identify all possible uses e.g., as input to
  other industries, plus direct consumption
• Establish technical norms of consumption for each
  end use
• Find out target levels of output in all
  industries consuming the product- including
  exports in targeted year for each industry
• Likely new developments involving product
• Add

8
Survey Methods

• Expert Opinion Poll
• Market consultants, industry analysts with
  knowledge of product and the market conditions
• Personal insights can be subjective
• To avoid the problem of dominant personality,
  Delphi method

9
Delphi method

• Developed by Rand Corporation in 1940s- arriving
  at consensus
• - Anonymity
• -Wide expertise
• -Effective when there is no urgency
• but
• -Difficulty in getting panelists
• -Requires understanding, skill and knowledge for
  conceptualising, stimulating discussion and
  making inferences by researcher .

10
Market Studies and Experiments

• Consumer Clinics or controlled lab experiments
  where consumer is given an amount and expenditure
  behaviour is observed
• Aware of being observed, consumer may not behave
  naturally
• Costly
• Not large enough to generalise

11
Market Experiments

• Market Experiments
• Changes are introduced in select markets and
  consumer response studied
• Gives time for producer to make necessary changes
  if necessary
• Costly errors are avoided

12
QUANTITATIVE METHODS

• 1. Time Series Method- naïve forecasting
• - Variables change with time
• Sources of variation in Time series
• Secular Trend
• Seasonal Changes
• Cyclical Fluctuations
• Random or irregular fluctuations

13
Time Series Method-

• Total variation , say in sales, is the result of
  all four factors operating together.

14
Time Series Method

• Trend Projection
• Simplest- projecting the past trend by fitting a
  straight line to the data either visually or more
  precisely through regression

15
Least Squares Method

• Least Squares Method- Most widely used time
  series method
• Linear Equation of a straight line is
• Y a bX
• where Y is the demand and X is the time period
  (no of years), a and b are constants depicting
  intercept and slope of the line. Calculation of Y
  for any value of X requires the values of a and
  b, for which 2 normal equations are prepared

16
Least Squares Method

• SY na b SX
• SXYa SX b SX2
• With values of a and b, straight line equation is
  obtained and forecast is made for Y for given
  value of X.

17
Time Series Method

• Smoothing Techniques
• These predict values of a time series on the
  basis of some average of its past values.
• Useful when time series exhibit little trend or
  seasonal variations but a great deal of
  irregular or random variations

18
Time Series Method

• Moving Average Method
• 3 (or 5) monthly/ yearly/ quarterly moving
  averages computed
• Average value of the last 3 (or 5) entries
  becomes the forecast for the next period

19
Time Series Method-

• Simple moving average gives equal weight to all
  observations, even though more recent
  observations are likely to be more important.
• Exponential smoothing overcomes this problem.

20
Barometric Forecast

• . Barometric Forecast
• When data indicates cyclical fluctuations
• To predict short term changes in economic
  activity or turning points
• Barometric forecasting is done by NBER and the
  Conference Board

21
Barometric Forecast

• Related variables are categorised into 3 groups-
• Leading variables those that change before the
  actual change
• Coincident Variables Change along with variable
• Lag variables Follow the event
• If leading variables are identified, easy to
  predict actual variables

22
Barometric Forecast
PEAK
B. Leading
A.Coincident variable
C. Lagging variable
Time
Trough
PEAK
23
Barometric Forecast

• Leading indicators
• Building permits, new private housing units
• Number of loan applications
• New orders for durable goods for their components
  and raw materials
• Index of consumer expectations
• Stock prices

24
Barometric Forecast

• Coincident indicators
• Rate of unemployment
• GDP
• Industrial production
• Manufacturing and trade sales

25
Barometric Forecast

• Lagging Indicators
• Commercial and industrial loans outstanding
• Change in consumer price index for services

26

• Commonly used Macroeconomic Predictive
  Indicators
• Hiring
• Consumer spending
• Consumer confidence
• Purchase managers index
• Bank Lending
• Shipping activity

27
Barometric Forecast

• Problems
• Only used for short term forecasting
• Difficult in identifying and getting data on
  variables

28

• Hiring, consumer spending, consumer confidence,
  purchase managers index, bank Lending, Shipping
  activity

29
Econometric Modelling for Forecasting

• Identifying and measuring the relationship
• Can be single variable or multivariate regression
  model
• Single equation models for a firms demand Large
  multiple equation models for the entire economy

30
Regression Analysis

• Steps in Regression Analysis
• 1.Identification of relevant explanatory/
  independent variables
• 2.Collection of data on variable under forecast
  and its determinants- can be time series or cross
  section
• 3.Specifying the appropriate demand function for
  Estimation (Linear, log linear etc)

31
Regression Analysis

• Dx a0 a1Y a2 Pxa3Pya4Ae
• a1,, a2 , a3, a4 respective partial regression
  coefficients- measure of elasticity- measure both
  magnitude and direction of change
• ? Error term shows effect of omitted variables
  or any error in measurement

32
Regression Analysis

• 4. Estimating the above function by using the
  collected data on the variable and its
  determinants to get the values of the
  coefficients as well as coefficient of
  determination, R2. Higher the R2 better the fit
• 5. Forecast for the variable, given the estimated
  values of coefficients e.g., you can forecast
  expected sales if you know the future Y, P, A P
  of substitute etc

33
Regression Analysis

• Not subjective like the qualitative methods
• Based on causal relationships and produces
  accurate results
• Method is consistent
• Forecasts both direction and magnitude of change
  BUT
• Uses complex calculations
• Costly and time consuming

34
Simultaneous Equations Method

• Involves specification of a number of economic
  relationships, one for each behavioral variable
  and its estimation.
• Method leads to a complete model which can
  explain the behavior of all the variables

35
Risks in Demand Forecasting

• Inadequate analysis of the market- include all
  potential users of a product
• Forecasting all drivers of market in each
  segment
• Unforeseen events
• Petroleum industry

36
Sums in Forecasting

• Data for demand for watches for 5 years is given,
  estimate demand for 2014
• Year 2005 2006 2007 2008 2009
• No 120 130 150 140 160

37
Sum in Forecasting
Year X Y X2 Y2 XY
2005 1 120 1 14400 120
2006 2 130 4 16900 260
2007 3 150 9 22500 450
2008 4 140 16 19600 560
2009 5 160 25 25600 800
Total 15(SX) n5 700 (SY) 55 (SX2 ) 99000 2190 (SXY)
38
Sums in Forecasting

• Normal equation Y a bX (i)
• SY na b SX (ii) SXY aSX b SX2
  (iii)
• 700 5a 15b (iv) 219015a 55b (v)
• Solving (iv) and (v) we get,
• 10 b 90 b9 Substituting the value of b in
  (iv)
• 7005a15 9 5a565 a113
• Y1139X. For year 2013, X will be 10.
• Y 2013 113 9 10
• 203 watches