Suitably Flip

Pre-Zarqawi Adjusted Presidential Approval Rating: 55%

Have you ever peeked behind the positive/negative number in the oft-trumpeted approval ratings?

Harris Interactive has a new poll out today (from a survey conducted between June 2 and June 5, before the demise of al Zarqawi), so let's take a closer look at what goes into the metric.

Here's how repsondents answered the question:  "How would you rate the job President Bush is doing: excellent, pretty good, only fair, or poor?"

Excellent:      10%
Pretty Good:  23%
Only Fair:      22%
Poor:             45%
Not Sure:        1%

From this breakdown, the pollsters combine "Excellent" and "Pretty Good" to come up with the "Positive" rating, and "Only Fair" and "Poor" to divine the "Negative" rating.

They don't actually ask, "Is your opinion of the President positive or negative?"  They use the above subjective classifications, then decide that abject ambivalence must fall somewhere between "Pretty Good" and "Only Fair".

Does "Only Fair" sound like an apt description of performance equally sub-neutral as "Pretty Good" is above neutral?  My own subjective understanding of "Only Fair" would be slightly above true ambivalence, perhaps synonymous with "Passable" or "Adequate".  Not a ringing endorsement, but quite possibly above the neutral line, and nowhere near the equivalent of "Pretty Bad", which symmetry in the responses would require.

Instead, if one assumes the real "Neutral" threshhold would more appropriately fall somewhere between "Only Fair" and "Poor", the implied positive rating jumps from 33% to 55%.  (If you think "Only Fair" is no better than ambivalence, then maybe you split the "Only Fair" category and wind up at an adjusted approval rating of only 44%.)

Either way, the point is the poll takes subjective evaluations (distinct in their relative order, but not in magnitude), and aggregates them into another subjective, binary classification that may or may not sync up with respondents' views.

If you stick the "Neutral" or ambivalence threshhold between the 2nd and 3rd response, as Harris does, you'd presumably expect to see equally strong positive and negative characterizations at the 2nd and 3rd, and at the 1st and 4th responses.  Subjective as the responses are, this objectively is not the case.  If they want to keep "Excellent" and "Pretty Good" above the ambivalence line, then "Pretty Bad" and "Terrible" would seem to be more appropriate counterparts to yield any reliable positive/negative measurement.

Yes, polls are just polls and subject to all kinds of error and interpretation.  But if a pollster wants to glean a positive/'negative rating, why not ask that question directly, unless the pollster wants to retain the flexibility to game the numbers?

For reference, below are some other notable pre-Zarqawi "Fair or Better" ratings:

Condoleezza Rice:  79%
Karl Rove:  58%
Donald Rumsfeld:  56%
Dick Cheney:  55%

Comments

Pre-Zarqawi Adjusted Presidential Approval Rating: 55%

Have you ever peeked behind the positive/negative number in the oft-trumpeted approval ratings?

Harris Interactive has a new poll out today (from a survey conducted between June 2 and June 5, before the demise of al Zarqawi), so let's take a closer look at what goes into the metric.

Here's how repsondents answered the question:  "How would you rate the job President Bush is doing: excellent, pretty good, only fair, or poor?"

Excellent:      10%
Pretty Good:  23%
Only Fair:      22%
Poor:             45%
Not Sure:        1%

From this breakdown, the pollsters combine "Excellent" and "Pretty Good" to come up with the "Positive" rating, and "Only Fair" and "Poor" to divine the "Negative" rating.

They don't actually ask, "Is your opinion of the President positive or negative?"  They use the above subjective classifications, then decide that abject ambivalence must fall somewhere between "Pretty Good" and "Only Fair".

Does "Only Fair" sound like an apt description of performance equally sub-neutral as "Pretty Good" is above neutral?  My own subjective understanding of "Only Fair" would be slightly above true ambivalence, perhaps synonymous with "Passable" or "Adequate".  Not a ringing endorsement, but quite possibly above the neutral line, and nowhere near the equivalent of "Pretty Bad", which symmetry in the responses would require.

Instead, if one assumes the real "Neutral" threshhold would more appropriately fall somewhere between "Only Fair" and "Poor", the implied positive rating jumps from 33% to 55%.  (If you think "Only Fair" is no better than ambivalence, then maybe you split the "Only Fair" category and wind up at an adjusted approval rating of only 44%.)

Either way, the point is the poll takes subjective evaluations (distinct in their relative order, but not in magnitude), and aggregates them into another subjective, binary classification that may or may not sync up with respondents' views.

If you stick the "Neutral" or ambivalence threshhold between the 2nd and 3rd response, as Harris does, you'd presumably expect to see equally strong positive and negative characterizations at the 2nd and 3rd, and at the 1st and 4th responses.  Subjective as the responses are, this objectively is not the case.  If they want to keep "Excellent" and "Pretty Good" above the ambivalence line, then "Pretty Bad" and "Terrible" would seem to be more appropriate counterparts to yield any reliable positive/negative measurement.

Yes, polls are just polls and subject to all kinds of error and interpretation.  But if a pollster wants to glean a positive/'negative rating, why not ask that question directly, unless the pollster wants to retain the flexibility to game the numbers?

For reference, below are some other notable pre-Zarqawi "Fair or Better" ratings:

Condoleezza Rice:  79%
Karl Rove:  58%
Donald Rumsfeld:  56%
Dick Cheney:  55%