CSO, 09 December 2024, 11am
The monthly Industrial Production and Turnover (IPT) survey monitors the volume of production and value of turnover of industrial local units. Two indices are compiled from the survey – the monthly index of production and the monthly index of turnover.
The monthly IPT survey covers all the larger local units which between them account for the bulk of industrial output for NACE divisions 05 to 35.
Seasonal adjustment is concerned with identifying the seasonal fluctuations and calendar effects within a time series and then subsequently removing these fluctuations from the data. Seasonally adjusted data adds value to users by providing clear insights into the evolution of the trend over time and by providing more meaningful period on period comparisons.
The raw data for the Irish industrial production index displays high levels of volatility where large monthly changes are sometimes recorded for the unadjusted series. This reflects the very high levels of globalisation in the Irish economy, and as such, monthly seasonal adjustment is complex.
In addition, a relatively small number of multinational enterprises have a dominant share of the national industrial production index. These multinational enterprises often tend to report on a quarterly basis and the data received can contain profit adjustments and fiscal reconciliations which lead to negative values being reported, particularly in the latter periods of the year. As a result of these types of accounting adjustments the cumulative value for the three months which make up a quarter will be as intended, but the structure of the monthly data within each quarter can be more volatile.
Therefore, in order to understand and manage this high degree of data volatility and to aid users in the usage of these statistics, the Central Statistics Office (CSO) has reviewed the seasonal adjustment procedure related to the Irish IPT. While this review evaluated the modelling of individual series, it also considered whether the existing Direct Approach should continue to be used or whether the Indirect Approach would be better suited. It should also be understood that changes over time are not unique to any one seasonal adjustment approach and are likely to persist regardless of what approach to seasonal adjustment is chosen.
Currently, the CSO has a policy of using the direct seasonal adjustment approach for all industrial production and turnover indices. This means that each individual index is assessed and modelled independently. A key benefit of this approach is that a more accurate model is created for high-level aggregate series. A disadvantage of the Direct Approach is that additivity amongst related series is not guaranteed, and additivity is increasingly unlikely in the event of unstable or volatile series.
The indirect seasonal adjustment method starts at lower-level aggregates and combines these to create overall seasonally adjusted indices. This ensures that additivity is achieved between the component series and the aggregate and avoids situations whereby contradictory results are potentially seen amongst closely related series. However, due to the volatility of the component series being used, a disadvantage of this approach is that the high-level aggregates may become less accurate and not reflect their actual underlying seasonality.
The review undertaken has concluded that the relative benefits of the direct versus the indirect methods are very marginal and there are challenges with either approach. The review has also concluded that there is not enough evidence to suggest that switching to an indirect method of seasonal adjustment would bring about improved results in the short term at this time. In addition, there is a longer time series of data available to generate the seasonal adjustment models using the Direct Approach compared with the Indirect Approach.
As a result, the CSO will continue to seasonally adjust the IPT data using the Direct Approach of seasonal adjustment. The IPT team will also continue to work with the CSO's seasonal adjustment experts each month, prior to the monthly release, to ensure that any significant monthly fluctuations are best accounted for.
The CSO will also conduct an annual review of the models used in full to ensure more accurate adjustments are being made. In the longer term, the CSO will consider whether a further review of the seasonal adjustment approach used in the IPT should be undertaken to ensure the best approach continues to be utilised.
The seasonal adjustment modelling was conducted in accordance with the European Statistical System’s (ESS) Guidelines on Seasonal Adjustment (2024) and the CSO’s policy on Seasonal Adjustment (2019). The CSO uses the semi-parametric X-13ARIMA-SEATS program to conduct seasonal adjustment. This program initially fits a regARIMA regression model to a time series, adjusting for outliers, trading-day effects and holiday effects, and then forecasts the series forward.
As part of the re-basing procedure, new seasonal adjustment models were developed for each of the re-based series. In general, the series were seasonally adjusted using the automatic seasonal adjustment procedures in-built in the seasonal adjustment software. However, critical, and high-profile series were further manually revised to ensure high-quality seasonal adjustment diagnostics. The CSO adopts a partial concurrent adjustment approach to seasonal adjustment. With this approach, ARIMA models, filters, outliers, and calendar regressors are re-identified once a year and then fixed for the year. With this method the parameters and seasonal factors are re-estimated and updated every time new or revised data becomes available.
As the seasonal adjustment is conducted using the direct seasonal adjustment approach, aggregate series are adjusted without reference to the component series and additivity between aggregate and sub-aggregate series can be lost. Therefore, to mitigate the potential loss of additivity, the CSO has tried to maintain as much consistency as possible in the models of the important aggregate and sub-aggregate series, without compromising the overall quality of the seasonal adjustment.
Learn about our data and confidentiality safeguards, and the steps we take to produce statistics that can be trusted by all.