1346.0.55.003 - Interpreting Time Series: Are you being misled by the Seasons, 2012  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 11/12/2012  First Issue
   Page tools: Print Print Page Print all pages in this productPrint All

WHY SEASONALLY ADJUST?

WHY SEASONALLY ADJUST?

Data collected regularly over time may display seasonal patterns. These patterns can make it difficult to see the effects of other influences on your data. Seasonal adjustment is a process for removing seasonal patterns. This process can help explain underlying activity in your data.


WHAT IS A SEASONAL PATTERN?

Regular highs (or lows) at certain times of the year indicate a seasonal pattern.

Figure 1: Example Seasonal Pattern and Trend

Figure 1: Example Seasonal Pattern and Trend


SOME CAUSES OF SEASONAL PATTERNS

Seasonal patterns are caused by regularly repeating cycles in the real world. For example:
  1. Retail sales are always larger in December due to Christmas.
  2. Crops are harvested in the same season each year.
  3. School and public holidays, especially during summer, are popular times to go travelling.

WHAT IF WE REMOVE SEASONAL PATTERNS?

Removing seasonal patterns can help you understand what has been happening to your data. This is because there is more to your data than just seasonal patterns. There is also some sort of underlying direction, known as the trend. Once you know the trend and the seasonal pattern everything left over is known as the residual noise or irregular part of the series.


KEY POINT:

The process of estimating and removing seasonal patterns is known as seasonal adjustment.


TELL ME MORE ABOUT THE TREND

The trend is the underlying direction of the series - "what’s normal". It smooths out most of the noise and short-term effects present in the seasonally adjusted. If the noise is strong then the seasonally adjusted data will be quite volatile, but the trend will give you a much better idea of the long term behaviour of the data.


KEY POINT:

The trend is estimated by smoothing noise out of the seasonally adjusted series.


WHY IS MY DATA NOISY?

Noisy, or volatile, data is caused by a number of factors:
  1. It may depend on what we’re looking at, for example, actual crop yields are naturally volatile due to factors such as the weather.
  2. It may also be related to the way data is collected, for example, estimates of the unemployment rate will contain further volatility if we don’t survey the entire population.
If there is lots of noise in a series, this can obscure any regular patterns, and make it very difficult to understand what has been happening in your data.

Figure 2: Example Residual Series

Figure 2: Example Residual Series


KEY POINT:

If your data is very noisy or has weak seasonal patterns it may be difficult to obtain reliable seasonally adjusted estimates.


IRREGULAR EVENTS

Sometimes once-off, irregular events will have an impact on your data. For example, a strike by pilots would lead to a sudden reduction in air travel. Irregular events can affect our ability to identify seasonal patterns, leading to potentially misleading seasonally adjusted estimates.

Irregular events must be corrected to obtain reliable estimates of the seasonal pattern.

Figure 3: Example Irregular Event

Figure 3: Example Irregular Event


KEY POINT:

The impact of irregular events and noise can be seen in the seasonally adjusted but are removed from the trend.

Seasonal patterns, irregular events and residual noise can all make it difficult to understand what has been happening to our data.

Seasonal adjustment is a process for separating these influences. Seasonally adjusted and trend estimates allow us to highlight underlying characteristics of data.