Dunno. I was trolling at the time, but the author seems to acknowledge that the data from the english emails are worthless.
That's absolutely not what that passage says, they are just talking about two different things. The english language emails clearly show that discriminatory bias exists, as every legislator can speak English and their results are highly significant. (p<0.01)
So right there the description of 'worthless' is thrown in the trash. It's super informative.
I see this crop up in a lot of threads. Every competent research paper has a part of its discussion where they talk about what the possible limitations of the research are. This is all an important part of science.
Then he seems to make a big deal about the spanish email responses, which i guess is all that he really has?
That was based on a 10% response rate of voter id supporters responding to anglo named spanish language email senders, and 0% to hispanic named spanish language email senders. I don't know how interesting or meaningful that is. I think more or less worthless was a pretty accurate description.
The spanish language part is more effective at looking at what the precise motivation for that might be. A few problems, first you seem to be looking at table 6 when you should be looking at table 4 for the differences in support among voter ID supporters.
As for the results, the paper actually goes into how interesting and meaningful that is. Using a 1 tailed difference of means test and an n of 936 you get a result significant at the P<0.01 level for voter ID supporters, which is highly significant.
I do agree that table 6 was highly interesting though, although for different reasons. What the author is focusing on there was that both Republican and Democratic legislators who OPPOSED voter ID laws showed quite small (and statistically insignificant) response biases. Republican legislators that favored voter ID laws showed a result significant at the p<.1 level. That helps account for party bias.
It's hard to see how you could come to your conclusion.