Using textual analysis, Baker, Bloom and Davis (2016) (hereafter BBD) have
developed an index to proxy movements in policy-related economic uncertainty
In doing so, the BBD index uses articles from major newspapers. 4,5
In an article, BBD scans for three groups of words related to 1) economy, 2)
uncertainty and 3) economic policy. There should be at least one word from
each group in the article for the article to be counted as an indicator of policy-
related economic uncertainty.
For example, if an article includes words from the
first group such as output, production, etc. And from the second group such as
uncertain, unclear, etc. But fails to include words from the third group such as
spending, deficit, regulation, budget, tax, policy, etc., then it is not counted as an
indicator of EPU. To construct the index value, BBD counts the number of
articles which pertain all three groups of words. The frequency of words that
belong to each of the three groups is not relevant for the index computation.
This number of articles is scaled by the total number of articles in the same
newspaper to get a normalised final index. Several robustness checks and
alternative and unrelated measures of EPU yielding similar results to BBD6 have
bolstered the confidence that this newspaper-based measure accurately
captures EPU. This being said, some shortcomings remain.