Why Economic Figures Change: Revisions, Seasonal Adjustment and Calendar Effects in Türkiye

[سفير طرابزون]

Why Does an Economic Figure Change After Publication?

You might save a report of production growth, return to the official table a month later, and find a different rate for the same period. Two headlines may even report growth and decline for the same month because one uses unadjusted data and the other a seasonally adjusted series. Before calling the figures contradictory, check the series definition, the data version and the comparison being made. This guide explains revisions, seasonal adjustment and calendar effects in Turkish economic time series, rather than providing another list of general indicators.

All numerical examples are hypothetical and teach comparison and calculation. The year 2026 refers to the original guide’s review on 8 October; it does not turn an example into an actual result for Türkiye’s economy. A complete figure needs a value, unit, period, definition and data version. Omitting any one can change the conclusion even if every number is copied correctly. We will use these details to prepare a report that readers can verify for themselves.

Istanbul Financial Center in an archive photograph from September 2023.​

1. Identify the series before interpreting its movement

Start with the source’s series name and code, then its unit and frequency. “Industrial production, monthly, index” differs from “sales, monthly, million lira”. Higher monetary sales may reflect prices, quantities or both, while a volume index uses its own methodology to separate price effects. The word “adjusted” does not by itself mean inflation has been removed: price adjustment and seasonal adjustment are different operations.

Record the geographical area, sectors covered and base year too. Manufacturing figures are not necessarily figures for all industry, and survey data may differ from administrative records. If an intermediary website puts two series on one chart, return to their original definitions before comparing them. A series code helps when names are translated or a website interface changes. It does not replace reading a methodological-change notice if the code stays the same while coverage changes.

2. Reference period, release date and data vintage

The reference period is the time described: a month, quarter, year or particular week-ending date. The release date is when the figures were published. The data vintage identifies the version available when you obtained the series, which may contain later revisions. In a hypothetical example, a May release describes March, then a June table revises March. The June figure is a newer estimate for March, rather than a result of June’s economic activity.

When documenting a story, state the month concerned and the release used. When comparing two reports, check whether the second writer used a newer vintage. Nor does “the latest figure” necessarily refer to the same period across indicators: one monthly series may extend to August and another only to July. Do not make their dates appear consistent by concealing this difference. Separate columns for the reference period and release date make a monitoring table understandable and help prevent old data recirculating under today’s headline.

3. Unadjusted data: activity as recorded

An original, or unadjusted, series contains seasonal patterns, working-day effects and irregular events alongside the underlying economic movement. It is not a “wrong” version: it answers a question about activity recorded within its definition. It is useful when describing an actual count or period total and may serve operational or accounting needs. Comparing a holiday month with a full working month, however, may not isolate the direction of activity.

Imagine a factory producing 200 units one month and 180 the next, when it operated on fewer days. The 10% fall in the total is real in this example. Concluding that capacity or demand fell by 10% requires examining operating days and conditions. A story about shipment volumes may need the unadjusted total; a story about the movement between months may be better served by the official adjusted series. Choose according to the question, and name the version rather than leaving that choice hidden.

4. Seasonality: recurring patterns requiring a defined method

Seasonality is a pattern recurring at particular times of year, such as a tourism season, harvest or spending cycle. Seasonal adjustment estimates and removes that component to improve comparisons between months or quarters with different usual patterns. The central bank’s study of seasonal adjustment in economic time series explains the concept and methods. It is a conceptual reference, rather than proof that every current series uses a particular software package.

The process does not remove every shock or turn a fluctuating series into a straight line. An unusual factory shutdown may remain in the adjusted figures because it is not a normal seasonal occurrence. Seasonal patterns also change over time; estimating them requires a sufficiently long history and statistical diagnostics. Do not delete high months from a chart and call the remainder seasonally adjusted. Use the official treatment, or disclose your method if conducting an independent analysis.

5. Calendar effects: working days and movable holidays

Two months may differ in length, working-day composition and holiday timing even if daily activity is unchanged. A holiday can also move between months or quarters from one year to the next. Calendar adjustment seeks to separate the effects specified in the series methodology. Check whether the version is calendar-adjusted only or both seasonally and calendar-adjusted; do not translate the two descriptions as though they were identical.

In a simplified example, 210 units over 21 working days and 200 over 20 both give an average of 10 units a day. Official adjustment is not necessarily a matter of dividing a total by the number of days: it models effects with regard to the activity concerned. A shop remaining open on a holiday may respond differently from a factory that closes. A single homemade adjustment therefore cannot replace the official series for every sector, nor should the effect of an additional day be assumed constant in all circumstances.

6. Match the comparison to the data version

Seasonally adjusted data can help with short-term month-on-month or quarter-on-quarter comparisons. For annual comparisons, a publisher may present unadjusted or calendar-adjusted data. Follow the table’s description: there is no rule allowing every annual figure to be relabelled adjusted or every monthly figure raw. Read the adjustment specified alongside the result, rather than relying only on a secondary headline.

Suppose the raw value rises from 100 to 110, while the adjusted value rises from 103 to 104. The first gives 10% growth and the second about 0.97%. The example does not show that either figure is false: their statistical treatments differ. Write “the unadjusted total rose” or “the seasonally adjusted reading improved”, according to the series used. They may even move in opposite directions when raw activity rises by less than its usual seasonal increase.

A quick comparison

Data version	What can it help you read?	What should you check?
Original, unadjusted	Activity recorded within the definition	Seasonal patterns and working days when comparing periods
Seasonally adjusted	Movement between periods with different seasonal patterns	Seasonal adjustment does not remove price effects
Calendar-adjusted	Day and holiday effects under the methodology	The precise treatment stated in the table
Latest revised vintage	The current estimate of the series’ history	Both comparison values come from the same vintage
An earlier release vintage	Information available at an earlier date	Its date when analysing a historical decision

The Skyland towers in Istanbul’s Seyrantepe district, in an archive photograph from May 2024.​

7. Why are economic data revised?

Additional records or survey responses can arrive after the first release; errors may be corrected, or weights, sources and methods updated. Some revisions arise when new observations change estimated seasonal patterns. TÜİK’s published revision policy provides a framework for revising statistics. Understanding a revision’s reason and scope is more useful than assuming a figure changed to suit the direction of a news story.

Distinguish a routine update to provisional data, a major revision rebuilding a longer history, and correction of a specific error. These are different processes and should not be collapsed into one sensational headline. Read the source’s announcement, then state what actually changed: levels, growth rates, weights or period coverage. The mere existence of a revision does not establish its quality; transparency, consistency and the ability to reconstruct the comparison after revision also matter.

8. The end of a series can change more than its distant past

Seasonal estimates at the end of a series have less information about subsequent periods because those periods have not yet occurred. As more months become available, the estimated value for the latest month or several recent months can change. Claims of a recent turning point therefore need proportionate caution. Follow the next release rather than building an elaborate account around a small movement in a single observation.

In a hypothetical example, the first release puts the previous month at 100 and the current month at 101, giving 1% growth. A newer vintage changes them to 101 and 101.5, giving about 0.50%. The calculation has not changed; its inputs have. Comparing the new 101.5 with the old 100 would produce 1.5%, a result belonging to neither internally consistent official vintage. Obtain both comparison values from the same release, even if your immediate aim is to update only one month.

9. Rebasing, breaks and classifications

Rescaling an index to a new base year is a matter of measurement scale; by itself it does not represent economic growth or decline. Changes to coverage, activity classification or data sources can raise wider comparability issues, however. When a series moves to a new base, look for the officially linked historical series. Do not combine levels from two different tables merely because the indicator names are similar.

Do not fill a methodological break with a guess and conceal it in a chart. If no consistent series is available, mark the change and define the period over which comparisons can be defended. Also check whether totals and components remain additive after adjustment. Some methods do not make the sum of separately adjusted components equal a total adjusted in another way. If they differ, examine direct and indirect adjustment before calling it a simple addition error.

10. Flows, stocks and averages: what can be added?

A flow measures activity during a period, while a stock measures a position at a particular time. Monthly exports, when consistently defined as flows, can be added to obtain a period total. End-of-month deposit balances cannot be added to produce “the year’s deposits”: the question may require the latest balance or an appropriate average. A weekly average also differs from a week-ending value.

Do not add index levels or growth rates as though they were monetary amounts. Nor should monthly data be converted to quarterly frequency without establishing whether the variable requires a sum, average or end-period observation. If changing units from thousands to millions, update every column consistently. These apparently small details can inflate a headline by a factor of a thousand or produce a meaningless rate. Keep the calculation with the table so readers can see how the original data led to the result.

11. A practical verification and archiving process

Start with the release, table and metadata, then consult the current year’s revision plan if TÜİK produces the series. Save a dated copy of the file used. In the next release, compare the same earlier period across vintages, recording the revision’s size and the reason if the publisher gives one. This separates new economic developments from changes in estimates of past activity.

Do not generalise one series’ policy to another. For example, the central bank’s Weekly Money and Banking Statistics metadata specify that the data are not seasonally adjusted and are updated when source data are revised. Do not assume that every central-bank figure includes seasonal adjustment. This is a check of a particular definition, rather than a reason to apply industrial-production or GDP rules to banking statistics.

12. Reading scenarios and writing a clear report

If the raw total grows while the adjusted series falls, activity may be higher but below its usual seasonal pattern. Explain both versions without using one to cancel out the other. If last month’s growth changes after revision, report the old and new rates from two documented releases, then calculate the new month’s growth using revised values. If two websites disagree, check their update times, units and series codes before concluding that the underlying sources conflict.

A report might read: “Seasonally and calendar-adjusted data showed a change from the preceding period, while the new release revised the preceding period’s reading.” Use “provisional” when the source does. Avoid “final figure” without support, and avoid “removing inflation” when discussing seasonality. Distinguish an analysis using the latest available historical series from an assessment of a past decision requiring the information available on that date. Using later revisions to judge earlier knowledge needs explicit explanation.

13. Frequently asked questions

Are adjusted data always more accurate than the originals?
Each version answers a different question. Unadjusted data help describe actual totals; adjusted data help analyse movement between periods with different seasonal patterns. The choice depends on purpose and definition.

Does seasonal adjustment remove inflation?
No. Separating seasonal patterns differs from separating price changes. Read the unit and calculation method to establish whether a series measures volume or value, rather than relying on the word “adjusted”.

Why revise a month already published?
New data may arrive, statistical estimates may change, or an error may be corrected. Check the stated reason and scope. A changed number alone does not establish a political or technical explanation.

Is a moving average the same as seasonal adjustment?
No. Averaging smooths fluctuations but may retain seasonal patterns and delay the appearance of turning points. Name the method and do not mix an average with an officially adjusted series.

Does comparing a month with the same month a year earlier remove seasonality?
It helps comparison, but does not guarantee removal of working-day effects, movable holidays or changing patterns. Use the source’s description when it provides a calendar-adjusted version.

Should I keep an old release after revision?
Yes, if documenting what was known when a report was written. Use a consistent newer vintage for current analysis; identify the vintage available at the time when evaluating a historical decision.

How can I identify a break in a series?
Read footnotes, change announcements and metadata. Rebasing alone does not necessarily imply a break; changing coverage may require a linked series or a new boundary for comparisons.

14. Checks before publishing a chart or rate

Record the series name, code, unit, frequency and coverage, together with its reference period, release and data vintage. Specify the adjustment accurately and ensure both comparison values come from the same vintage. Distinguish flows from stocks, levels from rates, and seasonality from price effects. Check revisions, breaks and weights; retain the file and calculation. Label hypothetical examples clearly, and do not make a small movement in one month a definitive judgement about the economic cycle. The original guide records its last review of references and methodology as 8 October 2026.

Related topics ​

• Retail and Consumer Spending
• Trade and Logistics
• Energy Prices
• Industry and Manufacturing
• Companies and Deals
• Borsa Istanbul, Gold and the Lira
• Banking and Monetary Policy
• Weekly Economic Indicators
• Using EVDS 3 Data