Page 400 - Contributed Paper Session (CPS) - Volume 4
P. 400
CPS2449 Louisa Nolan et al.
syntactic differences for identical products) make it difficult to automatically
clean the data to a structured state that is ready for aggregation and analysis.
Optimus is a natural language processing (NLP) pipeline that retrieves
vector representations of item descriptions and allows tiered grouping of both
syntactically (words that look the same) and semantically (words which are
contextually similar) similar descriptions. This produces a structured dataset
where each item can be classified across multiple hierarchical tiers. Data can
then be aggregated to an appropriate level or linked to existing taxonomies.
3. Results
All the faster indicators show promise as early-warning indicators of large
changes in the economy, although none, on their own, should be used as
proxies for gross domestic product (GDP) or other official economic statistics.
Figure 3 shows a comparison of the quarterly VAT turnover index with GDP
from 2008 to 2018. The index is strongly negative during the recession, but
there is a much more scattered relationship with GDP during periods of
greater stability. This demonstrates how novel uses of data can give a new
window on the economy (we may identify economic turning points more
quickly than waiting for official figures) but that care should be taken with
interpretation (the diffusion index is not a proxy for GDP).
The shipping indicators and road traffic indicators for HGVs are, as one
might expect, more closely correlated with trade in goods than with GDP, but
again, there is enough scatter in the relationship to make them unreliable as
a direct proxy, not least because we do not, at present, know what – if anything
– is being transported the ships or lorries. However, the geographical
granularity of these indicators is useful. Changes in activity in and around ports
is likely to have an economic impact, and if we can link this to the types of
products shipped at particular ports, we have the potential not only for an
early warning, but also perhaps some information about the types of industries
that might be affected.
ONS is now publishing the new faster indicators on a monthly basis (4).
They are published as research outputs, to reflect the fact that they are still in
development, whilst allowing users to explore them and offer feedback.
389 | I S I W S C 2 0 1 9