Pollster reader Dave asks:
Why have you stopped including summary numbers in your graphs?
As most of you know, our charts usually include a regression trend line (a line drawn through the points) as well as a legendd at the top of each chart that includes the most recent value of the trend estimate for each candidate. This example below shows the final candidate estimates for the Indiana Primary:
However, as Dave and other alert readers have noticed, the charts for West Virginia and all of the remaining primary and caucus states except Oregon no longer display any sort of "summary number." Their legends look like this:
The reason is that we do not plot a regression trend line when we have less than eight polls. Why not? Here is Professor Franklin's explanation posted when we started running these charts last year:
Ideally, the trend estimator should have a dozen or more polls to work with before we take the trend very seriously. When the number of polls drops too low, the trend estimator will jump around considerably if new polls are very far from previous polling and may produce jagged trend lines that are likely to change with more data. Despite this danger, we've estimated the trend with as few as eight polls, rather than stick to the safer minimum of twelve. Too many states are between 8 and 12 polls to ignore, and while we are cautious most of the trends look pretty reasonable even with less than 12 polls. When the polls are consistent with each other, the trend estimate will still be pretty good even with eight. But when there is substantial disagreement among the polls, we will get jagged or otherwise "bad" trends. Since the plots show the actual polls you can look at the data yourself and decide whether the trend is a reasonable fit to the data, or if it is erratic enough to be discounted. You decide.
Until this past weekend, the legend on charts with less than eight polls had displayed the candidate numbers for the median poll (or the average of the two middle polls when we had 2, 4 or 6 polls available. While that approach made sense to us last year, we found two practical problems: (a) the median poll was sometimes very outdated and (b) nearly everyone was confused by this approach. Many concluded the numbers in the box were clearly "wrong" as they didn't seem to match the dots on the chart. So rather than adding a confusing explanation in small print, we thought it better to err on the side of caution and simply drop the "summary numbers" when we have fewer than eight polls.
As of this morning, the number of available polls is just shy of the magic number 8 in West Virginia (6) and Kentucky (7) and a long way off in Montana (1), South Dakota (1) and Puerto Rico (1).
For those who must have a summary number, the average of the last four polls for West Virginia as of this writing (all conducted in the last two weeks) is Clinton 61%, Obama 24%