Demand Forecasting

1
Chapter 5

• Demand Forecasting

2

• 1.Importance of Forecasting
• Helps planning for long-term growth
• Helps in gauging the economic activity (auto
  sales, new home sales, electricity demand)
• Reduces risk and uncertainty in managerial
  decisions.

3
Types of Forecasts

• Qualitative Forecasts- Forecasts based on the
  survey of experienced managers
• Quantitative Forecasts- Forecasts based on
  statistical analysis (Trend projections)

4

• 2.Qualitative Forecasts
• Surveys and opinion polls
• executives and Sales persons. They are used to
• Make short-term forecasts when quantitative data
  are not available
• Supplement quantitative forecasts
• Forecast demand for new products for which data
  do not exist.

5

• 2Qualitative Forecasts Examples
• Surveys of business executives plant and
  equipment expenditure plans
• Surveys of plans for inventory change and
  expectations
• Surveys of consumers expenditure plans

6

• Opinion polls
• -Executive polling
• -Sales force polling
• -Consumer intention polling

7

• 4.Quantitative Forecast Methods
• Time Series Analysis - use of past values of an
  economic variable in order to predict its future
  value.
• Trend Projections (linear trend, growth rate
  trend).

8
Types of Time Series Data Fluctuations

• Secular trend-long-run upward moments or downward
  movements (population size, evolving tastes)
• Cyclical fluctuations-fashion, political
  elections, housing industry experiencing decline
  and rebounding)
• Seasonal Fluctuations- Housing starts, Hickory
  Farm sales Nov-January, Christmas sales

9

• Irregular or random fluctuations variation in
  data series due to unique events such as war,
  natural disaster, and strikes.

10

• 6. Trend Projection
• Extension of past changes in time series data
  into the future (sales, interest rate, stock
  value forecasting)
• a)Constant amount of change or growth
• Sales f(time trend)
• St a bt ? constant amount
• of growth

11

• b) Exponential growth function
• St So(1g)t constant percentage growth
  (exponential growth)

12

• 6a. Linear Trend Projection

13

• Demand for Electricity in KWH(million)
• Year St t Year St t
• 92-1 11 1 94-1 14 9
• -2 15 2 -2 18 10
• -3 12 3 -3 15 11
• -4 14 4 -4 17 12
• 93-1 12 5 95-1 15 13
• -2 17 6 -2 20 14
• -3 13 7 -3 16 15
• -4 16 8 -4 19 16

14

• St 11.90.394t R2.5
• S17 11.9 .394(17) 18.60
• S18 11.9 .394(18) 18.99
• S19 11.9 .394(19) 19.39
• S20 11.9 .394(20) 19.78

15

• 6b. Exponential Growth Projection
• Model St S0 ( 1 g)t
• ln St lnS0 t ln(1 g)
• Year lnSt t
• 92.1 2.398 1
• . . .
• . . .
• . . .
• 95.4 2.944 16

16

• ln St 2.49 .026t
• Taking the antilog of both sides yields,
• St 12.06(1.026)t R2 .5
• S17 12.06(1.026)17 18.76
• S18 12.06(1.026)18 19.14
• S19 12.06(1.026)19 19.64
• S20 12.06(1.026)20 20.15

17

• Notice that forecasts based on linear trend model
  tend to be less accurate the further one
  forecasts into the future.

18

• 7.Methods of Incorporating Seasonal Variation
• Ratio to trend method
• Group the data by quarters
• Get a forecasted value for each quarter by using
  the trend model
• Calculate the actual/forecast ratio for each
  season or each month.
• Find the average of the actual/forecast ratio for
  each season over the entire period of the study.

19

• b. The dummy variable method
• Multiply each unadjusted forecasted value of the
  economic variable by its corresponding seasonal
  adjusting factor.
• Include n-1 dummy variables in the trend equation
  and run the regression.

20

• Time-Series Growth Patterns

Y
Y
Y
Time(t)
Time(t)
Time(t)
(b)Exponential growth trend
(c)Declining rate of growth trend
(a)Linear trend
21

• 8.Some shortcomings of Time Series Analysis
• Assumes that past behavior will be repeated in
  the future
• Cannot forecast turning points
• Does not examine the underlying causes of
  fluctuations in economic variables.

22

• 9.Smoothing Techniques (Irregular Time Series
  Data)
• Refer to the methods of predicting future values
  of a time series on the basis of an average of
  its past values only
• They are used when the data show irregular
  variation (random).

23

• Moving Averages
• Help to generate acceptable future period value
  of a variable when the time series are subject to
  random fluctuations.
• -See, Table 5-5 in the handout
• 3-quarter vs 5-quarter Moving Average Forecasts
  and Comparison
• Objective Forecast 13th quarter value,
• given time series data for the previous 12
  quarters
  

24

• Choose the appropriate period based on the
  lowest RMSE.
• RMSE
• At actual value of the time series in period
  t.
• Ft the forecasted value of the time series in
  period t.
• Problem Gives equal weight to
• each period

25

• b. Exponential smoothing
• - a smoothing technique in which the forecast
  for period t1 is a weighted average of the
  actual (At)and forecasted values(Ft) of the time
  series in period t.

26

• Ft1 wAt (1-w)Ft
• where Ft1 the forecast of F in period t 1.
• w the weight assigned to the
• actual value of the time
• series, 0ltwlt1.
• 1-w the weight assigned to
• the forecasted value of
• the time series.

27

• 10. Using Econometric Models to Forecast
• Advantages
• Seek to explain the economic phenomenon being
  forecasted- i.e. enables mgt to assess the impact
  of changes in policies (price, Ad)
• Predict the direction and magnitude of change

28

• Models can be modified based on the comparison of
  actual and forecast value.
• Examples
• Comment The above advantages have to be weighed
  against the difficulties of getting the forecast
  values of each of the explanatory variables.