Follow Paul Donovan

Polling problems are more than political

| Posted by: Paul Donovan | Tags: Paul Donovan Weekly

  • Opinion polls did not have a good UK election. Most polls clustered around a 7% to 8% Conservative lead. The result was 2.3% lead. There were good reasons for the surprise (including a high youth turnout), but opinion polls demonstrated errors once again.
  • Why do economists care? Because nearly all economic data is survey-based, just like opinion polls. Survey errors in economic data have been getting larger and more frequent in recent years, resulting in significant data revisions.
  • Errors in surveys are a particular risk with business confidence surveys. These are just opinion polls with clever branding. Even when surveys ask "factual" questions like, "have export order volumes increased in the last month?" the response
    is still likely to be an inaccurate opinion – "I think" rather than "I know."
  • Survey data (especially in the US) is more volatile than the underlying economy. Survey data has overreacted to real economic performance. Thus surveys that correct from current strong but exaggerated levels may not indicate an economic downturn, just a return to reality. The US ISM and the Euro area PMI measures should start to moderate, but it would be unwise to assume that the moderation automatically predicts a slowing of economic activity.