Attibution and President Bush’s approval rating December 7, 2008
Posted by Paul Duignan in : Impact evaluation, Attribution, Reporting systems, Outcomes systems architecture, Indicators, Outcomes theory & the news, Accountability, Measurement , trackback
I’ve been doing a lot of thinking about indicators and attribution these days. I think that a lot of problems (particularly misinterpretations of how well or badly parties are doing within outcomes systems of all types arise because of confusion between attributable and not-necessarily attributable indicators. The underying Five Building Blocks Diagram which is used in outcomes theory makes a distiction between attributable and not-necessarily attributable indicators.The Washington Post today had an interactive showing President Bush’s Approval Rating over time. At various points, the graph had labels showing what was happening at the time, presumably to help the reader work out why his rating went up and down at the points it did. The biggest jump was around September 11 where his approval took a massive leap from something like 56% to 92% (reading off the graph). Below is a section from the interactive.
Source: Section from Washington Post Interactive 7 December 2008 (http://www.washingtonpost.com/wp-dyn/content/custom/2006/02/02/CU2006020201345.html).
Now an interesting question is to what extent the jump in approval is attributable to President Bush? By attributable here I mean that we are able to establish that what changed the indicator (we will call it an indicator if that is OK for our purposes here) was something that President Bush did. This implies that if President Bush had acted in a different way, or been a different person, the indicator may not have changed as much as it did. (Have a look at the article on attribution and other features of steps, outcomes and indicators).
One hypothesis is that it was the way that President Bush handled the emergency that lead to the jump in his approval rating. The other is that the approval rating of any president would have made a massive jump in these circumstances. Without further analysis there’s no way of knowing how much of each of these components affected the rating. I don’t know if an analysis of crises, how presidents have handled them and what has happened to their approval ratings would be able to throw much light on this situation or not.
What we are left with from this is that the jump in President Bush’s approval rating, in terms of attribution to President Bush, should be regarded as a not-necessarily attributable indicator - one for which we cannot actually establish attribution back to him. By it being a not-necessarily attributable indicator, this does not mean that it was not influenced by what he did in reality, it is just that we cannot prove that he influenced it at the moment (and may not ever be able to prove this).
Now the whole point about this from a technical outcomes theory point of view leads us to wonder what type of indicator we normally think we are dealing with when we think about the President’s approval rating? I suspect that initially most people (myself included), would think of it as an attributable indicator - after all it is his approval rating isn’t it? Why would it not be attributable to him?
But, as shown above, if we think about it in more depth it seems that it is a not-necessarily attributable indicator at least at some points in time. I will talk in a later posting about what this different perspective on the President’s approval rating may mean.
Paul Duignan, PhD
Comments»
no comments yet - be the first?