Reply #178: An additional response to your specific question

[Febble]

which I didn't address. You asked: "I wonder why this graph bothers you so?"

It bothers me for two main reasons. The first is that the labels that indicate the voting technology are simply wrong. Most states used a mixture of technologies, but that wouldn't matter if the label represented, say, the dominant technology, or if there were ANY logic to the division between "electronic" and "paper" categories. But there isn't. Anyone can check that, but no-one seems to.

The second reason is that their provenance is unknown. I once found them on Wikipedia attributed to - me. I deleted them. I'm a scientist and an academic, and the first rule of academia is that you cite your sources. This data is unsourced and demonstrably erroneous, yet it still keeps being linked to.

Now you seem to know the source - if so, perhaps you could inform whoever made the plots that the labels are in error. As exit poll plots alone they are moderately useful. As a set of plots that appear to be assembled to make some point about the relationship between exit poll discrepancies and voting technology, they are simply misleading. In fact, the biggest exit poll discrepancies were not associated with electronic voting at all but with lever machines. This is of interest, but for reasons I can only speculate about, I have seen no interest from anybody in finding out why the exit poll discrepancies should have been most strongly associated with levers than with other technologies (punchcards came second). New York, which had a large exit poll discrepancy, a discrepancy calculated from the actual lever counts at the precincts (not central tabulations), still retains its levers. Its voters like them. They don't like DREs (rightly). So why doesn't this set of plots not show the exit poll discrepancy for New York, in a plot marked "levers"? I can only guess because it doesn't fit the narrative that the election was stolen electronically and the evidence lies in the exit polls. But that narrative is not supported by the data.

Lastly - you mentioned that the authors of the graph used "info {from} BEFORE THE SCRUB". And so they should. The adjusted exit poll numbers don't tell you the discrepancy because, of course, they were adjusted. But it would be wrong to describe the unadjusted data as "before the SCRUB" firstly, because it reflects a complete misunderstanding of the adjustment process (post-stratification reweighting), secondly, it displays an apparent unawareness of the fact that it is, and remains, standard procedure in polls (i.e. it was not some unprecedented cover-up), and, thirdly, implies that the unweighted data was permanently "scrubbed" from the record, when in fact not only was it freely available for download at the time, as it was again in 2006, but was published in great detail by the pollsters themselves (more accurately in fact, as the un-"scrubbed" data was already partially re-weighted).

As I think justice is important, and as I think accurate data is important, I make these points.

Cheers

Lizzie