Seasonal effects are not base effects
as illustrated by euro area inflation
Business cycle analysis is largely done using time series. However, these series not only contain information about the business cycle. Throughout the year, for instance, changes also occur that have to be attributed to changing seasons or more or less fixed socio-economic patterns. It is therefore best to use series that are seasonally adjusted. Year-on-year comparisons are a possible alternative to circumvent seasonal effects. However, they don’t give a sharp picture of the most recent dynamics as they may be biased by base effects. The latter played a major role in the evolution of euro area inflation in 2024. Eurostat inflation figures are not seasonally adjusted, which can give a blurred, sometimes misleading picture of real, underlying inflation dynamics. Therefore, seasonally-adjusted series published by the European Central Bank (ECB) are best used to analyse euro area inflation.
Business cycle and time series analysis
Economic analysis is very often based on time series. These show the development over time of an economic variable, such as gross domestic product, unemployment or the general price level, with a fixed periodicity, for example annually, quarterly, monthly or even weekly.
The analysis of time series thus allows the study of the evolution through time of the economy. This is often more difficult than it may seem at first glance. Indeed, the economy is constantly changing and these movements can be the result of both random as well as fundamental, economic factors.1 Trend factors, such as demographics or technology, cause relatively moderate economic changes that only stand out when considered over a longer period of time. On top of trend changes, an economy typically exhibits cyclical fluctuations. This so-called business cycle change can be driven by many factors, such as a demand shortfall, capacity constraints, changes in economic policy or external events such as a sudden sharp change in the price of energy or other commodities or simply a loss of confidence.
Business cycle analysis aims to capture this short- and medium-term economic evolution. It thereby provides important information for investment decisions by investors and firms, and for economic policy. It is regularly and sometimes extensively reported in the press. Not infrequently, it is also the subject of debate because the business cycle picture is not always clear-cut. After all, the time series on which the business cycle analysis is based are also determined by movements that are not necessarily related to the business cycle.
Thus, there are fluctuations throughout the year that recur periodically at more or less fixed times and with more or less constant intensity. Seasons are typical examples. In winter, there are usually more weather-related lost days in construction, resulting in lower production there. However, this does not mean that construction goes through a recession every winter. Some legal regulations or social traditions also influence economic activity with a more or less fixed annual pattern. Just think of company closures due to annual breaks or legal holidays.
These so-called seasonal fluctuations can fortunately be statistically eliminated and therefore do not have to interfere with a sharp business cycle picture. In contrast, other factors such as an exceptionally harsh winter, natural disasters, major strikes or one-off socio-economic policy adjustments have no fixed pattern and remain in the time series even after seasonal adjustment. Thus, seasonally adjusted time series also contain changes that are not necessarily related to business cycle movement, but rather reflect random short-term volatility. It creates noise in the business cycle picture.
Seasonal adjustment versus year-on-year comparison
Business cycle analysis prefers to use seasonally adjusted time series. Remaining noise can then be smoothed out by working with a three-month moving average of the series. Random fluctuations from month to month are thus somewhat disregarded and comparing successive three-month periods is probably the best way to monitor the economy.
But sometimes seasonally adjusted series are not available. Then year-on-year change approach is often used. Here, the figure for the latest available period is compared with the same period a year earlier. The dynamics of the indicator under consideration is then estimated on the basis of its percentage increase or decrease against the previous year.
Small intermediate fluctuations from month to month, whether seasonal or not, mostly have limited impact in this approach. Compared to the short-term dynamics that can be measured via a seasonally adjusted series, a year-on-year change gives a much less precise picture of recent dynamics. Indeed, the year-on-year change rate can be the result of both a recent change in dynamics as well as a change a year ago. If, for example, the number series under consideration experienced a sharp drop (increase) a year ago, the rate of change will increase (decrease) today, even though nothing has changed in dynamics in the recent month. This so-called base effect - the influence of the rate of change by the (comparator) base and not the recent development - obviously makes for an inaccurate measurement of the current economic pulse at the time of measurement. After all, it is distorted by the pulse of a year ago.
Illustration: energy price inflation in the euro area
The sometimes incredibly large impact of short-term volatility on year-on-year rates of change can be strikingly illustrated by looking at the course of inflation in the euro area in 2024, more specifically at its energy price component. Inflation is traditionally measured by comparing the price index in the month under consideration with the price index a year before, in other words as a year-on-year comparison.
Figure 1 shows the course of the energy price component in the harmonised consumer price index for the euro area in 2024, as well as the course in 2023, advanced one year. The latter means that the months of 2023 were advanced to the corresponding months of 2024. The figure illustrates the erratic course of energy prices, especially in 2023.
On balance, energy prices fell by almost 7% during 2023: the index fell from over 157 in December 2022 to less than 147 in December 2023. In 2024 too, energy prices still fluctuated quite sharply, but less than in 2023 and on balance they remained almost stable: in December 2024, the index stood at 147, about the same as a year earlier (146.9). Nevertheless, over the whole of 2024, energy price inflation was on average sharply negative (over -2%), fluctuating from -5% in January to +1.3% in July to again -6% in September and +0.1% in December, as illustrated at the bottom of Figure 1.
The figure also illustrates how the annual changes during 2024 were determined much more by the large fluctuations in the basis of comparison - the index in 2023 - than by the changes in 2024 itself. For example, the sharply negative year-on-year changes in early 2024 was caused by sharp fall in energy prices in early 2023, which completely neutralised the slight rise (versus the previous month) in energy prices in January and February 2024. The sharp fall in energy price inflation in August and September 2024 was primarily caused by a temporary rebound in energy prices in the corresponding months of 2023 and only to a lesser extent by the fall (versus the previous month) in energy prices in August and September 2024.
Inflation has seasonal pattern
Apart from energy prices, food prices can also be very volatile at times. To find out the underlying inflation dynamics, economists often look at core inflation. This is the annual increase in the consumer price index, excluding energy and food prices. In principle, this core inflation has a more stable trend.
Nevertheless, it should not be forgotten that the consumer price indexes published by Eurostat for the euro area are not seasonally adjusted, while price adjustments do show a seasonal pattern.2
The seasonal pattern of core inflation for the euro area is illustrated in Figure 2 by the month-on-month changes of the index in each month of the year over the past decade. Certainly in the five-year period before the outbreak of the pandemic - a period of relatively stable but below the ECB's 2% targeted inflation - the differences in price adjustments in the respective months over the years were very limited. But the price adjustments did vary widely from month to month every year, to about the same extent. This indicates a seasonal pattern. In January, for instance, prices traditionally fell, while they rose sharply in March. Only in the months of April to June did the size of price adjustments vary relatively more from year to year. This is partly due to the fact that Easter and other holidays do not fall in the same month every year. Price adjustments for package holidays, for example, therefore do not fall in the same month every year.
Since the pandemic, monthly price adjustments have varied more from year to year than before. This makes sense given the numerous shocks the economy has experienced since then. Initially, lockdowns prevented prices for some goods and services from being set at all. This created a predominantly downward effect on inflation. The reopening of the economy was accompanied by catch-up movements on price adjustments, but also by major disruptions in supply chains, which also interfered with the normal pricing pattern. This then resulted in a predominantly upward inflationary effect. This was compounded by the direct pass-through of high food and energy prices into core inflation, e.g. in prices for meals in a restaurant. It was subsequently reinforced by its second-round effects via wage adjustments. These create stronger price increases to this day than before the pandemic. Yet it is also striking that despite the greater variation in monthly price adjustments in the years since the pandemic, the pre-pandemic seasonal pattern has remained largely intact.
The seasonal pattern of price adjustments implies that assessing short-term inflation dynamics on the basis of non-seasonally adjusted monthly figures can easily lead to wrong conclusions. Nevertheless, a correct assessment of those short-term inflation dynamics is essential for a central bank with the achievement of an inflation target as its main mandate, such as the ECB.
Perhaps not coincidentally, therefore, the ECB itself publishes seasonally adjusted series of the harmonized consumer price index for the euro area. These allow for a more accurate mapping of underlying inflation dynamics via the appropriate three-month moving average technique, as Figure 3 illustrates. The seasonally adjusted ECB figures suggest a gradual cooling of short-term core inflation dynamics to 1.8% annualised in January 2025. The Eurostat figures suggest a much sharper fall in inflation, but most likely wrongly. Indeed, they fully consider the usual annual drop in prices in January relative to December as a sign of cooling momentum, ignoring the fact that prices are very likely to rise again in February and especially March, as they do every year. With a probability bordering on certainty, according to this measurement, the so-called inflation dynamics will rise again in the coming months, when in fact essentially seasonal factors will be at work.
Similarly, an outlook on the likely path of inflation due to likely base effects gives a misleading picture when seasonal factors are not taken into account. Figure 4 illustrates this using the expected path of core euro area inflation should no price adjustments occur between February 2025 and January 2026. By definition, core inflation would then be 0% in January 2026. In between, the inflation rate is only determined by price changes during 2024, which serve as a basis for comparison. The non-seasonally adjusted Eurostat figures thereby show an inflation path that assumes no seasonal price adjustments in 2025. Thus, they immediately show a sharp decline in the months of February to March. However, it can be assumed with high probability that seasonal price increases will also happen in these months in 2025. Therefore, the outlook based on the seasonally adjusted ECB series gives a better picture of the likely base effects. After all, seasonal effects are not (real) base effects!
1 For a more detailed explanation see: OECD, Data and Metadata Reporting and Presentation Handbook, 2007.
2 The consumer price index published by the US Bureau of Labor Statistics for the US is seasonally adjusted, though.