Lesson 5 of 7 · 15 min
Deseasonalising data
Remove the seasonal effect from a time series by dividing by the seasonal indices, reverse the process, and use deseasonalised values to compare seasons fairly.
What you will be able to do
- Calculate a deseasonalised value as actual value ÷ seasonal index.
- Find an actual value from a deseasonalised value: actual = deseasonalised × seasonal index.
- Use deseasonalised values to decide whether a change is more than the usual seasonal movement.
Guest nights at the holiday park fell from 41 thousand in Q1 2023 to 23 thousand in Q2 2023 — a drop of 44%. Should the manager panic?
Of course not: Q2 is always quiet. The real question is whether Q2 2023 was quieter than a Q2 usually is. To answer it, take the seasonal effect out of each value. That is deseasonalising.
Key idea
Deseasonalising
The deseasonalised (seasonally adjusted) value is what the value would have been if that season were an average one. Busy seasons (index above 1) come down; quiet seasons (index below 1) go up.
To go back: .
Worked example
Deseasonalise a year of guest nights
The holiday park's seasonal indices are Q1 1.262, Q2 0.698, Q3 0.964 and Q4 1.076. In 2023 it sold 41, 23, 31 and 35 thousand guest nights in Q1 to Q4.
a) Deseasonalise the 2023 values, correct to two decimal places.
b) What do the deseasonalised values show that the actual values don't?
Step 1: Divide each value by its own quarter's index
Q1: Q2:
Q3: Q4:
Step 2: Check the direction
Q1 and Q4 (indices above 1) came down; Q2 and Q3 (below 1) went up. ✓
A quick sense check that you divided rather than multiplied.
Step 3: Interpret
The actual values swing from 23 to 41 thousand, but the deseasonalised values are all close to 32.5 thousand. Almost all the swing is the usual seasonal pattern. Relative to a typical year, Q2 was the strongest quarter of 2023 and Q3 the weakest.
Try it yourself first, then check the working one step at a time.
Answer
a) 32.49, 32.95, 32.16, 32.53 thousand. b) Once the seasonal effect is removed, business was steady at about 32.5 thousand guest nights a quarter throughout 2023; Q2 was relatively the best quarter, even though it had the fewest actual guest nights.
Method
Divide each actual value by the seasonal index of its season.
Check that peak seasons came down and quiet seasons went up.
Compare deseasonalised values to judge underlying performance.
Watch out for
Multiplying by the index instead of dividing.
Using the wrong season's index for a value.
Green squares: the deseasonalised series. The big seasonal zigzag is gone, and what's left is a steady climb with small irregular bumps. The widget keeps every decimal of the indices, so its values may differ from hand calculations in the last place.
The holiday park sold 34 thousand guest nights in Q3 2024. The Q3 seasonal index is 0.964. Calculate the deseasonalised value, in thousands, correct to two decimal places.
| Row | A | B | C | D |
|---|---|---|---|---|
| 1 | Quarter | Guest nights ('000) | Seasonal index | Deseasonalised |
| 2 | Q1 2022 | 38 | 1.262 | =B2/C2 → 30.11 |
| 3 | Q2 2022 | 21 | 0.698 | =B3/C3 → 30.09 |
| 4 | Q3 2022 | 29 | 0.964 | =B4/C4 → 30.08 |
| 5 | Q4 2022 | 32 | 1.076 | =B5/C5 → 29.74 |
| ⋮ | ⋮ | ⋮ | ⋮ | ⋮ |
The deseasonalised value for a Q1 quarter is 36 thousand guest nights. The Q1 seasonal index is 1.262. How many guest nights were actually sold, in thousands, correct to one decimal place?
Deseasonalised values let you compare any two periods fairly, whatever their seasons. In 2024 the park sold 44 thousand in Q1 and 38 thousand in Q4. Deseasonalised, Q1 is and Q4 is . Relative to what each quarter normally does, Q4 was the stronger quarter.
The deseasonalised guest nights were 32.49 thousand in Q1 2023 and 32.95 thousand in Q2 2023, although actual guest nights fell from 41 to 23 thousand. What is the best conclusion?
Worked example
Did sales really rise? Comparing quarters with deseasonalised values
A hardware store in Mackay has seasonal indices of 0.94 for Q3 and 1.12 for Q4. Its sales were $386 000 in Q3 and $455 000 in Q4.
a) The owner says Q4 was a much better quarter than Q3. Use deseasonalised sales to evaluate this claim.
b) What Q4 sales would have been needed for deseasonalised sales to equal those of Q3?
Step 1: Deseasonalise Q3
Step 2: Deseasonalise Q4
Step 3: Evaluate the claim
Actual sales rose by $69 000, but deseasonalised sales fell by about $4400. Q4 is always a busier quarter (index 1.12), and the rise was smaller than the usual seasonal increase. The claim is not supported: allowing for the season, Q4 was slightly weaker than Q3.
Deseasonalised values compare each quarter with what that quarter normally does.
Step 4: Reverse the process
To put the seasonal effect back, multiply by the index.
Try it yourself first, then check the working one step at a time.
Answer
a) Deseasonalised sales were about $410 600 in Q3 and $406 300 in Q4, so after allowing for the season Q4 was slightly worse, not better. b) About $459 900.
Method
Deseasonalise both quarters before comparing them.
Compare the deseasonalised values and state the conclusion in context.
To find an actual value, multiply a deseasonalised value by the index.
Watch out for
Comparing actual values of different seasons and calling it a real improvement.
In (b), dividing by 1.12 again instead of multiplying.
Common mistake
Multiplying instead of dividing
Quick check: a peak season has an index above 1, so its deseasonalised value must be smaller than the actual value. If your Q1 value went up, you multiplied. The same check works in reverse for a quiet season.
History
Seasonally adjusted figures in the news
When the Australian Bureau of Statistics reports that unemployment "rose" or retail turnover "fell" in a month, it usually quotes seasonally adjusted figures. Without adjustment, every January would show a rise in unemployment (school leavers looking for work) and every December a boom in retail, telling you nothing new.
Summary
Deseasonalised value = actual ÷ seasonal index. Actual = deseasonalised × seasonal index.
Peak seasons come down when deseasonalised; quiet seasons go up.
Deseasonalised values show the underlying level — the trend and the irregular fluctuations — with the seasonal effect removed.
Compare periods from different seasons with their deseasonalised values, not their actual values.
Finished the lesson?
Try the 3 quick checks above and the lesson completes itself, or mark it complete now.