By Timothy Boobier
Seasonal adjustment aims to identify, estimate and remove regular seasonal fluctuations and typical calendar effects (e.g. numbers of trading days in a month) from time series data. This article describes the 2015 annual review and summarises its results. It also provides an update on other seasonal adjustment workstreams.
Annual Review Process
The Bank of England reviews the seasonal adjustment of published series on a regular basis. The frequency of reviews of particular series will vary, based on the usage of the series. The annual review is divided into three phases spread across the year, as detailed in Table 1.
For each data series reviewed, the following issues are routinely considered:
- presence of seasonality;
- seasonal adjustment settings:
- choice of ARIMA model;
- calendar effects1;
- seasonal and trend filters;
- residual seasonality; and
- direct versus indirect adjustment (selected series only). Some series can be adjusted either directly or indirectly (as the sum, or the difference, of their seasonally adjusted components). The method chosen depends on various properties of the series and its components, and the relationships between them. The Bank continues to monitor the adjustment structures of key series to ensure that these remain appropriate.
Results of 2015 review
In total, 138 published series were reviewed in 2015, as detailed in Table 2. This resulted in changes to 35 existing published series (Table 3).
One series is being newly published on a seasonally adjusted basis as a result of this review: PNFCs’ all currency net equity issuance was found to be seasonal and a seasonally adjusted series for this is being published with the publication of April 2016 data. This series is CPMZID5. This was not implemented earlier alongside Phase 2 implementation, due to further investigation of conceptual and methodological issues around seasonally adjusting capital issuance series. Other net capital issuance series used as part of the published net finance raised measures for PNFCs (e.g. net bond issuance) continue to be reviewed regularly for seasonality.
In addition to the regular review process, the Bank also conducts specific work on issues relating to seasonal adjustment:
- The seasonal adjustment of Cash ISA deposits with the household sector was affected by the introduction of new ISA rules in July 2014. This is the series LPMB4F6. The new rules included increased thresholds and allowed transfers from Stocks and Shares ISAs to Cash ISAs.7 These changes introduced a seasonal pattern for 2014 that was materially different from the typical seasonal pattern observed in previous years. As a result, the Bank introduced a new regressor for this effect, implemented for the September 2015 data publication. This consisted of a stock regressor constructed with reference to differences between 2014 data points and historical averages. For months showing a material difference, those cumulative differences were incorporated into a stock regressor.
- Following the switch to X-13ARIMA-SEATS in June 2014, the Bank reviewed alternative regressors for Easter effects on amounts outstanding series, which allowed the testing of different lags, compared to the one-day lag of the Bank’s user-defined regressor. In X-13ARIMA-SEATS, testing ‘easterstock’ regressors allows reviewers to choose between easterstock, easterstock, easterstock and no Easter. Changes to these regressors were implemented for relevant series for the April 2015 data publication, and the Bank opted to incorporate these regressors into its reviews going forward.
- In addition to this, the Bank is reviewing alternative regressors for trading day effects for amounts outstanding series, using the programme’s in-built trading day stock regressors.
- The seasonal adjustment of notes and coin series is under review due to changes in patterns of trading day and holiday effects, particularly from 2013 onwards.
To view all related tables, please download the full article:
For questions relating to this article please contact firstname.lastname@example.org or call +44 (0)20 3461 5356.