Seasonal adjustment models part of regular rebasing
Every year, SA’s economic activity peaks in the fourth quarter (OctoberDecember). On the production (or industry) side of GDP, this pattern is driven mainly by manufacturing; trade, catering and accommodation; and transport, storage and communication.
Underlying the upward surge in these industries is the festive season, which we see clearly on the expenditure side of GDP, with an annual peak in household expenditure in the fourth quarter.
On the production side, manufacturers know the big spending is coming, and ramp up production sharply in October and November, helping lift fourth-quarter GDP. In the first quarter (January-March) there is a corresponding GDP dip. As these and other seasonal patterns are relatively regular, their effect on GDP can be estimated and extracted from data to arrive at a “seasonally adjusted” GDP (and seasonally adjusted agriculture, mining and manufacturing figures). Showing economic growth excluding seasonal effects is important in gauging quarteron-quarter GDP performance.
In her recent column , Neva Makgetla raised negative firstquarter growth in GDP figures Stats SA published for the last four years since 2016 (“Skew first-quarter GDP data needs straightening”, September 30). She argues that Stats SA’s seasonal adjustment model is not doing enough, resulting in negative first-quarter growth rates that are not a true reflection of the economy.
The pattern of economic activity does change over time and Makgetla is right to point out the importance of keeping estimation methods up to date. Distinguishing between changes that are structural, seasonal, short term, long term or simply ad hoc are important in estimating GDP and the vast range of economic indicators that are used in its compilation.
Stats SA rebases and benchmarks national accounts every five years; the next round is due in September 2020; our seasonal adjustment models are very much part of this exercise. It can take years to diagnose which changes are seasonal and which are not.
An interesting question is which industries were behind the negative first-quarter growth of the past few years (quarter on quarter, seasonally adjusted). In 2016, mining was the largest negative contributor. In 2017, the main culprits were manufacturing and trade, both positive contributors in 2016. In 2018, the weakness was widespread, with negative contributions from agriculture, mining, manufacturing and trade. In 2019 it was these four again, but also transport, storage and accommodation. This mix of results is evident at lower levels. Of the 10 divisions in manufacturing production, two were consistently negative in the first quarter in the four years 2016-2019 (quarter-on-quarter, seasonally adjusted). The other eight were a mixture of positives and negatives. The more varied the results after seasonal adjustment, the trickier it becomes to say if the seasonal adjustment model requires an update.
Makgetla argues that the annualisation of growth rates can be misleading. There are options for the headline GDP growth rate, such as year-onyear using unadjusted values, year-on-year using seasonally adjusted values, or quarter-onquarter (not annualised) using seasonally adjusted values.
Stats SA has treated the seasonally adjusted annualised rate (production side) as the headline number for many years, but we are open to listening to our user community and considering all proposals.
WE ARE OPEN TO LISTENING TO OUR USER COMMUNITY AND CONSIDERING ALL PROPOSALS
Manamela is chief director of national accounts at Stats SA.