One characteristic feature of central banks today is that policy decisions are almost
exclusively made by a committee rather than by a single policy maker. Another is that
central banks endeavour to be transparent. Together, this has brought to the fore an
important but so far unresolved issue: how much information about the committee’s
deliberations should the central bank reveal? Does this kind of information make
monetary policy easier to understand and predict, or does it make it harder? We address this issue by employing a novel method. We measure the sentiment and tone of the minutes of the Swedish central bank using an automated content analysis that
converts the qualitative information in the minutes to a quantitative measure. We find
that this measure is helpful in predicting future policy rate decisions.
