ACCA SBL Syllabus G. Finance In Planning And Decision-Making - Time Series Analysis - Notes ^{8 / 8}

Any data collected over time (eg sales volumes) can be used here

Time series forecasting methods are based on the assumption that past patterns in data, such as seasonality, can be used to forecast future data points.

Such patterns allow for more curved than linear predictions.

Let’s take a simple example.

Over the past 6 years, a particular company has noticed that on month 12 the sales are usually 30% higher than typical monthly volumes.

Thus it makes sense to forecast that month 12 for the forthcoming year will follow a similar pattern

This graph shows a scenario where linear regression has predicted an increase in sales of roughly €4M per quarter

However Time series has taken into account past trends which suggest that Q1 sales are usually €4M below trend, Q2 are €4M above and Q3 are €4M below.

In time series analysis, the trend line itself may also be curved.

Indeed it would only be linear as the above example, if the favourable and adverse seasonal affects cancel each other out

• Time Series Analysis Components

  Time Series Analysis is made up of three main components used in different ways to produce future forecasts:

  • Average 

    the mean of the observations over time

  • Trend 

    a gradual increase or decrease in the average over time

  • Seasonal influence

    predictable short-term cycling behaviour due to time of day, week, month, year, season and so on

• Forecast data might also be affected by cyclical movement (unpredictable long-term cycling behaviour due to the business cycle or product/service lifecycle) and random error (remaining variation that cannot be explained by the other four components)

• Variations of time series analysis

  Time Series Analysis offers 2 main variations:

  • Moving Averages

    The forecast is based on an arithmetic average of a given number of past data points. 

    This should make the trend become more obvious.

    Let us take a simple example by considering the following data:

• Period	1	2	3	4	5	6	7	8	9	10	11	12
  Sales €M	47	50	51	48	48	52	52	49	50	52	54	50

• It is difficult to immediately spot the trend as the figures appear to be constantly increasing and decreasing.

  However, a moving average (average sales from periods 1-4, 2-5, 3-6 etc) of this data using 4 period averaging would give the following result.

• Moving Average

  49.00

  49.25

  49.75

  50.00

  50.25

  50.75

  50.75

  51.25

  51.50

• Exponential Smoothing

  • A type of weighted moving average that allows inclusion of trends etc. This gives greater weighting to more recent data in order to reflect the more recent trend.

    An exponential smoothing (average calculated by taking 4 times the 4th period, 3 times the 3rd period, 2 times the 2nd period and 1 times the 1st period and then dividing by a total of 10) of the data would present a similar picture

• Exponential Smoothing

  49.20

  48.80

  49.90

  50.80

  50.40

  50.30

  50.80

  52.10

  51.60

• Advantages and Disadvantages

• Advantages	Disadvantages
  Identifies seasonal variations	Complicated
  Can be non-linear	‘Seasons’ may change
  Accurate	Based on historical data
  	Less useful in the long term

Conclusion

Linear regression is most relevant when there is a linear relationship between the variables. 

On the other hand, time series analysis is most appropriate when seasonal variations causes curved forecasts.

The reliability of a forecasting method can be established over time. 

If the forecasts used, turn out to be inaccurate, management might decide to use alternative methods of forecasting.

On the other hand, if forecasts prove to be accurate and successful, providing management with key data for decision making, then it is more likely that management will continue to use the same forecasting methods.